chore(ci): diagnose open handles in smoketests

[jrvb-rl]

Description

Enable Jest open-handle diagnostics in the HTTP/1 and HTTP/2 smoketest jobs so the next normal dev/prod runs can identify resources preventing clean shutdown. This only gathers evidence; it does not change SDK cleanup or attempt a leak fix.

Motivation

All four transport jobs in the September 21 dev/prod runs passed their tests but reported a worker that could not exit gracefully and had to be terminated.

Changes

• Run smoketests with --runInBand --detectOpenHandles instead of --maxWorkers=800%.
• Print allocation stacks in the existing Actions logs, separately for each transport.
• Retain failure behavior and the existing 60-minute job timeout; do not add --forceExit or ignore errors.

Jest detection requires serial execution and adds overhead. It can change reproduction of concurrency-sensitive leaks. If a handle remains open, the job may reach its timeout after printing diagnostics. This is diagnostic configuration to revisit after collecting evidence, not a permanent leak fix.

Testing

• Prettier and git diff --check.
• Smoketest discovery with the diagnostic flags: 28 test files listed without executing live tests.
• Local synthetic timer-leak test: Jest printed the open-handle type and allocation stack; an external 10-second timeout stopped the deliberately leaking probe as expected.
• No dev/prod smoketests triggered.

Breaking Changes

None to the SDK API. CI smoketests run serially while diagnostics are enabled.

Checklist

• Conventional commit title
• Smoke test workflow updated
• Diagnostic behavior validated locally

[reflex-loop]

✅ Reflex agent status: Completed

Agent	Status
agt_2atOMtgR8vwhUg100l9xQI	✅ Completed

This PR was created by Reflex.

_{This comment updates in place as agents work.}

[jrvb-rl]

Let's add a TODO to remind us to bump up workers again once we've diagnosed the shutdown paths.

[jrvb-rl]

Added a TODO beside the diagnostic command to restore --maxWorkers=800% and remove serial open-handle diagnostics once the shutdown paths have been diagnosed.