fix(bcode-laminar): track pending flushes so shutdown awaits in-flight OTLP exports

[Alezander9]

rc5 diagnostic confirmed root cause: session.idle's await processor.forceFlush() never returns because the in-flight OTLP HTTP export is killed by process.exit before it can resolve. The host doesn't track these Promises (they're orphan microtasks).Fix: bus event handlers now call trackFlush(processor.forceFlush()) which stores each Promise in a module-local pendingFlushes Set. The sync shutdown hook awaits Promise.all(pendingFlushes) before returning, so the host's 3s race actually waits for the in-flight HTTP request to complete.This is the actual fix the previous five rcs were chasing.

---

Summary by cubic

Ensure bcode-laminar waits for in-flight OTLP exports on shutdown so spans aren’t dropped. We now track flush Promises from bus handlers and await them within the host’s 3s shutdown window.

• Bug Fixes
  • Track every processor.forceFlush() from bus events in a pendingFlushes Set via trackFlush.
  • Shutdown ends open spans, triggers a final forceFlush, then awaits Promise.all(pendingFlushes) (still bounded by the host race).
  • Improved stderr logs with counts and timings to confirm the path executed.

^{Written for commit 08d37e4. Summary will update on new commits. Review in cubic}