test(did-webvh): regression tests for same-second versionTime collision (#600)

[stormer78]

Root cause

Investigating the reported bug — offline vta services didcomm enable updates config but the VTA's did:webvh document keeps advertising REST only (versionId stays 1-) — traced to the same-second versionTime collision, not an offline-vs-online split.

• The offline path (services_cli::run_services_didcomm_enable → enable_didcomm → update_did_webvh) does reach the WebVH LogEntry publish step; it has since the op was written. It is not "returns after only writing config" — update_did_webvh runs first, then config.toml is written.
• The genesis entry (from vta setup) and the enable's update entry are minted in the same wall-clock second. did:webvh serialises versionTime at second granularity and requires each entry to be strictly later than the previous, so entry 2 collides. config.toml still gets written because the library stamps the entry without validating strict-increase at build time — the failure only surfaces when a client resolves did.jsonl, which truncates the log back to the version-1 REST-only state.
• This affects both transports (online pnm services didcomm enable and offline vta …) — any create-then-update in the same second.
• The double-quoted mediator_url from the report is not reproducible on current code: resolve_mediator extracts the endpoint via ServiceEndpoint::get_uri() (clean string). It points at a double-encoded serviceEndpoint on the mediator side, not a VTA config-writer bug.

Already fixed on main

The bug is already fixed by PR #600 (c73d9db, "backdate + space the VTA's webvh versionTime"), which landed after v0.10.11 (the build in the report). The user should upgrade to ≥ the build containing #600 (main is now 0.10.19). #600 backdates + index-spaces versionTime on both create_did_webvh (genesis) and update_did_webvh, via backdated_version_time.

What this PR adds

#600 shipped with only manual end-to-end validation and no automated regression test. This closes that gap:

• create_then_didcomm_enable_back_to_back_resolves (did_webvh e2e): drives create → didcomm-enable with no sleep in between (every other e2e test in the module dodges the bug by sleeping past the second boundary), then asserts the chain validates and advertises DIDCommMessaging at version 2. Temporarily neutering backdated_version_time makes it fail with the exact didwebvh-rs error: Current versionTime (...) must be greater than previous versionTime (...). Log truncated at 2-... — proving it's a genuine guard, not a vacuous pass.
• backdated_version_time_is_past_and_strictly_increasing (unit): pins the helper's contract — timestamps are in the past and strictly increasing by entry index at second precision.

Reuses the shared with_didcomm_service patcher rather than hand-rolling a service-array edit (per workspace CLAUDE.md).

Validation

• cargo fmt clean
• cargo clippy -p vta-service --all-targets — no new warnings (2 pre-existing, unrelated, in provision_integration/preconditions.rs)
• cargo test -p vta-service — all green
• Bumps vta-service 0.10.18 → 0.10.19

https://claude.ai/code/session_01HoRff8BDqXuSfQ7RNjm3ic