Modify ForceFlush to abort after timeout/cancellation

[humivo]

Modified the method, BatchSpanProcessor.ForceFlush, to abort after configured timeout or cancellation. Should resolve Issue #1618

[linux-foundation-easycla]

The committers are authorized under a signed CLA.

• ✅ humivo (a5db348, 98ae772, 54e1a6d, 23908b4, 65cf934, 3a432e9)

[codecov]

Codecov Report

Merging #1757 (93e2ec4) into main (3947cab) will increase coverage by 0.0%.
The diff coverage is 83.3%.

@@          Coverage Diff          @@
##            main   #1757   +/-   ##
=====================================
  Coverage   78.6%   78.6%           
=====================================
  Files        133     133           
  Lines       7072    7083   +11     
=====================================
+ Hits        5560    5569    +9     
- Misses      1267    1268    +1     
- Partials     245     246    +1     

Impacted Files	Coverage Δ
sdk/trace/batch_span_processor.go	75.0% <83.3%> (+0.6%)	⬆️

[MrAlias]

To test if ForceFlush succeeds it should be verified that if actually flushes any held span batches not just returns a nil error. Can you extend this test to do that?

[Aneurysm9]

--- FAIL: TestBatchSpanProcessorForceFlushSucceeds (0.00s)
batch_span_processor_test.go:292: number of exported span: got 516, want 517

This has been a recurring issue with attempting to test this span processor. Even though this test has set the queue size to 0 and the processor is configured to block on a full queue, it looks like it's still possible for the last span to be pulled off the queue and the scheduler to switch back to the test and start ForceFlush before processQueue finishes adding the last span to the batch.

I wonder if testing that a non-zero number of spans were exported or that the exported count was within 10 of the expected count might be sufficient here since we simply want to know that ForceFlush shipped spans and the remaining spans can be picked up later. @MrAlias or @open-telemetry/go-approvers any thoughts?