Move the samples to preview 2.0.301.4181 and adopt what it added

[romanett]

Proposed changes

Moves the samples from 2.0.280.50326-preview (stack commit c4960b54) to
2.0.301.4181-preview (stack commit 105508bc), 21 commits later, and adopts what that
range added.

Breaking changes the bump forces

The GDS packages lost their .Common suffix
(UA-.NETStandard#4438).
OPCFoundation.NetStandard.Opc.Ua.Gds.{Client,Server}.Common are now
…Gds.{Client,Server}, and so are the assemblies inside them — which the GDS sample
applications were already using as their own AssemblyName. MSBuild reported the collision
as MSB3243 and picked a winner at random, not as an error, so the two apps and the
controls library are renamed to Quickstarts.GlobalDiscovery{Client,Server,ClientControls}.
Their RootNamespace and every using are untouched; the sample catalog addresses them by
project path, so nothing else moves.

Part 11 Historical Access was completed
(#4390). Every historian
provider interface changed shape: collections are ArrayOf<T>, an update answers a
HistorianUpdateOutcome<T> (one status per entry plus the values it displaced, for the audit
trail), DeleteRawAsync answers one of those rather than a bare StatusCode,
InsertBatchAsync takes ArrayOf<HistorianDataBatch>, ReadAtTimeAsync answers
ArrayOf<DataValue>, and a resume token carries a ByteString. The three sample providers
follow. On the client, ReadRawAsync and friends gained a HistoryReadNodeOptions
parameter before the cancellation token, so a positional ct no longer compiles.

The fluent configure/seal path became awaitable
(#4442).
NodeManagerBuilder.Configure(…) and Seal() are gone; the two hand-written fluent
managers now run the pipeline the generator emits — CreateFluentBuilder, Configure,
RegisterAuthoredNodesAsync, CompleteConfigureAsync, SealConfigurationAsync.

Patterns adopted

• Historian capabilities are a gate, not only an advertisement. The dispatcher refuses an
  operation the provider does not claim, and the diagnostics node manager now rolls the event
  flags up as well. WellReportHistorianProvider claims the five event operations it serves,
  TestDataHistorianProvider claims the server timestamps its archive keeps, and
  HistoricalEventsCapabilities no longer sets five flags by hand — only the Server object's
  EventNotifier is left to it.
• HistorianEventRecord.QualifiedFields is filled from the select clauses beside the flat
  Fields, so two clauses which differ only in their event type no longer collapse onto one
  browse path. The sample's own BuildOperandKey gives way to HistorianEventFieldKey.BuildPath.
• NodeManagement mints through the NodeId factory
  (#4433) rather than a New
  override, which the AddNodes path no longer reaches — a node built from a TypeDefinition
  arrives carrying the type's id, so the override's !node.NodeId.IsNull guard silently sent
  every request to the base. NodeIdFactory.WithMode(NodeIdAssignmentMode.String) gives the
  sample back the readable identifiers it wanted.
• RoleManagement stores one wildcard endpoint filter instead of copying every advertised
  endpoint description. AddEndpoint accepts a partial EndpointType now
  (#4412), and Part 18 4.4.2
  says the fields left at their default are ignored anyway — so
  { SecurityMode = SignAndEncrypt } is the whole rule, and it keeps matching when a host
  name is spelled differently.
• The NodeSet2 InputArguments repair is deleted
  (#4422): the importer binds
  the typed property itself, so BindInputArguments and the README section explaining it are
  gone.

One real bug the bump exposed

DataTypesClientModel called LoadAsync(true, true) on the complex type system, and the
first parameter of that method is onlyEnumTypes. The client built no structured type at
all, so every structure the server serves arrived as an undecoded extension object — in the
sample whose entire point is that "a value of a structure the client has no compiled knowledge
of arrives decoded into its fields".

Nothing caught it, because the tier 1.7 fixture asserted against WrappedValue.ToString()
and the generated default value used to arrive as the bare XML of the model design
(#4401): the raw text happened
to contain Trek, Cube and 10, so the assertions held on a value that decoded into
nothing. #4414 fixed that generator defect, the body became a binary ByteString, and the
fixture started reporting Byte[35]. Both arguments are named at the call site now.

Recorded issues which paid out

Three KnownIssue wrappers failed because what they recorded now works, and are replaced by
plain assertions: the two structure default values in DataTypesNodeManagerTests (#4401,
fixed by #4414) and the audit trail in RoleManagementNodeManagerTests (fixed by
#4420). The prose describing
those gaps across four READMEs, docs/TESTING.md and three sources is updated.

Two expectations stay ignored because they are still real: a compiled type cannot hand over
its own DataTypeDefinition
(#4424) and no audit event of
an AlarmCondition condition-Method call reaches a subscriber. The second note is narrowed —
the RoleSet Methods of the RoleManagement sample are audited on this stack, so "no audit
event of any type reaches a subscriber" is no longer true.

Related Issues

Nothing to close here. What this bump unblocks or newly exposes is filed rather than crammed
in:

• Adopt the NodeId factory in the remaining sample node managers #875 — adopt the NodeId factory in the ten sample node managers which still override New
• Finish the fluent node manager migration: the hooks that blocked it have landed #876 — finish the fluent node manager migration; every SDK hook that blocked it has landed
• Historian samples: adopt the completed Part 11 provider surface #877 — demonstrate the completed Part 11 provider surface (old values, structured data,
  event history from the client)
• No sample overlays a NodeSet2 document through the fluent builder #878 — no sample overlays a NodeSet2 document through the new fluent Import

Types of changes

• Bugfix (non-breaking change which fixes an issue)
• Enhancement (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected, requires version increase of Nuget packages)
• Documentation Update (if none of the other choices apply)

The breaking box is ticked for the GDS assembly renames: anything naming
Opc.Ua.Gds.Client.dll or Opc.Ua.Gds.Server.dll from these sample projects by file name —
a binding redirect, a launch script — has to follow.

Checklist

• I have read the CONTRIBUTING doc.
• I have signed the CLA.
• I ran tests locally with my changes, all passed.
• I fixed all failing tests in the CI pipelines.
• I fixed all introduced issues with CodeQL and LGTM.
• I have added tests that prove my fix is effective or that my feature works and increased code coverage.
• I have added necessary documentation (if appropriate).
• Any dependent changes have been merged and published in downstream modules.

Further comments

Validation

UA Samples.slnx builds clean on net48, net8.0 and net10.0. Every tier was run locally on
net10.0:

Tier	Result
SampleConfiguration.Tests	237 passed
SampleServers.Tests	121 passed
SampleNodeManagers.Tests	181 passed, 1 skipped
SampleClientModels.Tests	421 passed, 3 skipped
SampleClients.Tests (WinForms)	65 passed

1,025 passed, 0 failed, 4 skipped. The four skips are the recorded known issues named
above plus one pre-existing skip unrelated to this change. The build carries no new warnings,
and the MSB3243 assembly conflict the rename fixes is gone.

On resolving the version

GitHub Packages needs a token with read:packages even for a public feed, which the local
gh token does not carry. The version was resolved and the build validated the way the
repository documentation describes for that situation: from the release-manifest.json of
the last successful nuget-publish.yml run on the stack's master, whose artifact doubles
as a local folder feed. Nothing about that workflow is committed — Nuget.Config is
untouched and targets.props changes only in the pinned version.

🤖 Generated with Claude Code