🧪 test(storage): share one capture subscriber

[gaborbernat]

test_webhook_queue_scan_skips_and_cleans_damaged_rows failed on some runs and not others, a different one of its three cases each time, on an unchanged tree. The count of the cleanup warning came back as zero, and the scan does emit it on every pass that finds damage, so something dropped the event on the way rather than the scan skipping it.

A callsite decides its interest the first time any thread executes it, and every thread reads that decision from then on. 🔍 While one dispatcher is alive, tracing resolves that question against the registering thread's own default rather than against the registry, so a thread holding no subscriber resolves it against nothing, caches Interest::never(), and silences the callsite for whichever case was asserting on it. A subscriber that lives only as long as one test is absent whenever a neighbour asks first, and which case loses that race is what varied between runs.

One subscriber for the whole binary cannot be the one missing at that moment. Events route to a capture belonging to the thread that raised them, so the cases still read only their own output while sharing the decision each callsite makes once. Serialising them would have hidden the interleaving rather than removing it, and it would leave the next test that captures logs in this crate to rediscover the same thing.

The assertion holds as it was, one warning across two scans, which is what proves the second scan found nothing left to clean. Nothing in WebhookQueueCleanup::log changed.

The new test drives the interleaving rather than waiting for it to come around. A thread holding no subscriber reaches a callsite first, and the event a subscribed thread raises at that callsite still reaches its capture. Swapping the shared subscriber back for a per-test one makes that capture come back empty, which is the failure this fixes.

Closes #2188

[codspeed-hq]

Merging this PR will not alter performance

✅ 29 untouched benchmarks
⏩ 133 skipped benchmarks¹

---

_{Comparing fix/webhook-capture-flake-2188 (0984a82) with main (d75b16a)}

Footnotes

1. 133 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩