fix(sdk-logs): avoid null _currentExport access in _flushAll

[Janealter]

Which problem is this PR solving?

BatchLogRecordProcessor._flushAll can throw TypeError: Cannot read properties of null (reading 'exportCompleted') when a forceFlush() call
races against the natural completion of a background export.

Reproduction

1. The processor has a background export in flight (started by the
   scheduled-delay timer or because a batch filled up). _exportOneBatch
   set this._currentExport = exportOp and attached
   exportOp.exportCompleted.then(() => { this._currentExport = null; … })
   as a fire-and-forget completion handler.
2. forceFlush() is called and enters _flushAll(). The check
   if (this._currentExport !== null) passes, and the code awaits
   this._exporter.forceFlush().
3. While that await is pending, the in-flight export completes and the
   .then handler above runs, setting this._currentExport = null.
4. _flushAll() resumes and evaluates
   await this._currentExport.exportCompleted. this._currentExport is now
   null, throwing TypeError.

In browsers this surfaces as an unhandled rejection because the browser
BatchLogRecordProcessor registers visibilitychange/pagehide listeners
that call void this.forceFlush() (no .catch() attached). Users
tab-switching during normal usage trigger the race very frequently in
production.

The same race is reachable from _shutdown() (which also calls
_flushAll()).

Affected versions

Introduced by #6356 (feat(sdk-logs)!: invoke exporter forceFlush without first awaiting export), shipped in @opentelemetry/sdk-logs@0.215.0. The
same code path is unchanged on main (verified against 0.215.0 through
0.218.0 and HEAD).

Short description of the changes

• Capture this._currentExport into a local inFlight const before the
  awaits, then dereference the local. The trailing
  this._currentExport = null is kept so the field still reflects "no
  in-flight" once the wait is done (idempotent if the completion handler
  already nulled it).
• Added a unit test in test/common/export/BatchLogRecordProcessor.test.ts
  (should not throw when in-flight export completes between awaits in _flushAll) that deterministically reproduces the race using a custom
  exporter with controllable export and forceFlush resolutions.

Without the fix, the new test fails with
TypeError: Cannot read properties of null (reading 'exportCompleted')
at the exact line affected.

Type of change

• Bug fix (non-breaking change which fixes an issue)

How Has This Been Tested?

• New unit test added that deterministically interleaves a forceFlush()
  call with the completion of a background export. Test fails on main
  with the exact upstream error, passes with the fix.
• All existing @opentelemetry/sdk-logs common tests still pass
  (nx run @opentelemetry/sdk-logs:test — 198 passing).
• nx run @opentelemetry/sdk-logs:lint passes.

Checklist:

• Followed the style guidelines of this project
• Unit tests have been added
• CHANGELOG entry added under experimental/CHANGELOG.md → Unreleased →
  Bug Fixes

[linux-foundation-easycla]

The committers listed above are authorized under a signed CLA.

• ✅ login: Janealter / name: Janealter (6759b88)

[pichlermarc]

good catch!
thank you for fixing this. 🙌

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.92%. Comparing base (8edcb66) to head (6759b88).
⚠️ Report is 3 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #6763   +/-   ##
=======================================
  Coverage   94.92%   94.92%           
=======================================
  Files         377      377           
  Lines       12818    12819    +1     
  Branches     2924     2924           
=======================================
+ Hits        12168    12169    +1     
  Misses        650      650           

Files with missing lines	Coverage Δ
...sdk-logs/src/export/BatchLogRecordProcessorBase.ts	91.21% <100.00%> (+0.05%)	⬆️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[otelbot]

Thank you for your contribution @Janealter! 🎉 We would like to hear from you about your experience contributing to OpenTelemetry by taking a few minutes to fill out this survey.