feat(tui): browse, index, and mute dataset sources from a Discover tab

[aaronstevenwhite]

Summary

Add a way to view available dataset sources from the TUI and select datasets to
index, with the public Layers PDS (repo.layers.pub) as a built-in default. That
PDS is deliberately kept off the firehose (so a small operator running thousands
of accounts is not blacklisted), which makes it otherwise undiscoverable; a
configured source fixes that, and the TUI grows a Discover tab, launch
auto-indexing, and permanent muting with review.

Changes

• Configured dataset sources. A Source model and a sources.toml config
  (under the XDG config directory, overridable with LAIRS_SOURCES_FILE) name the
  PDS and relay endpoints lairs crawls. A built-in default ships for
  repo.layers.pub; a user can add sources or override or disable a built-in.
  lairs index build --source <name> and lairs sources list use them.
• Discover tab (4). Browses the configured sources, crawls a chosen source
  in the background, and lists each dataset with its state (indexed, new, or
  muted). enter/space indexes a new dataset (it then shows on Explore) or
  permanently mutes an indexed one; r re-crawls. The crawl reuses the same
  listRepos path as lairs index build through a new streaming discover
  function in lairs.discovery.
• Launch auto-indexing. With no --index, the TUI opens a default index
  (under the XDG state directory, overridable with LAIRS_INDEX_DIR) and crawls
  the enabled sources on launch, indexing every newly discovered, non-muted
  dataset so Explore fills in on its own. lairs tui --no-auto-index skips it.
• Permanent muting with review. A muted dataset is recorded as a
  MutedDataset and excluded from auto-indexing until unmuted. A settings modal
  (ctrl+s) lists the sources and every muted dataset with an unmute action.
• Concurrency. Background workers only do network reads; every index write
  happens on the UI thread against one shared DiscoveryIndex, so there is never
  a concurrent writer. Explore reloads from disk on tab activation and after the
  auto-index commit.
• Tests. Unit coverage with a faked PdsClient (discover happy/skip paths,
  mute lifecycle, auto-index respects mutes, unreachable-source and crawl-failure
  handling, the Discover/Settings/Explore-reload flows, and the space/r
  bindings), plus live @pytest.mark.integration coverage against the dockerized
  test PDS: real crawl provenance, mute-respecting re-crawl, the full app
  end-to-end over a real PdsClient (listRepos -> describeRepo ->
  listRecords -> index -> Explore), and a multi-page drain at the default page
  size (205 corpora across 3 pages, no drops or duplicates).
• Docs. The explorer guide and API reference document the Discover tab,
  sources, settings, and auto-indexing.

Type of change

• Bug fix (a non-breaking change that fixes an issue)
• New feature (a non-breaking change that adds functionality)
• Breaking change (a fix or feature that changes existing behaviour)
• Documentation
• Refactor, test, or chore (no user-visible change)

Checklist

• uv run ruff format --check lairs tests passes
• uv run ruff check lairs tests passes (including ANN401; no Any in annotations)
• uv run ty check --error-on-warning passes
• uv run pytest passes (the unit and functional suite)
• uv run pytest --run-integration -m integration passes against the Docker test PDS
• uv run --group docs mkdocs build --strict passes (when docs or a public API changed)
• Tests added or updated for new behaviour, mirrored 1:1 under tests/
• CHANGELOG.md updated for user-visible changes
• No @dataclass, pydantic, Any, or object annotation introduced; no em-dashes
• Commits follow Conventional Commits, imperative mood, with no AI-assistant mentions

Related issues

Follow-on to the discovery index and cross-account materialization work; does not
close a specific tracked issue.