Add pool poison recovery integration coverage

[lodyai]

Summary

Milestone 4 added direct unit coverage for client::pool::sync::lock_or_recover, but the deeper integration contract should be tested separately: a panic during pool scheduler or slot bookkeeping recovery should not permanently corrupt subsequent scheduler behaviour.

Acceptance criteria

• Add an integration or module-level test that simulates poison during a waiter-table or slot-bookkeeping update.
• Recover through lock_or_recover.
• Run the next scheduler or slot operation and assert consistency is re-established: waiter counts remain coherent, fairness order is preserved where applicable, and no waiter is lost.
• Keep the test focused on pool bookkeeping; do not apply poison recovery to connection, protocol, lifecycle, serializer, or user state.

Context

This was identified during the CodeRabbit review for PR #534 while implementing the code-base audit plan in docs/execplans/code-base-audit-2026-06-05.md.

[coderabbitai]

🔗 Related PRs

leynos/rstest-bdd#496 - Surface GPUI scenario names on panic (10.1.4) [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!