ConcurrentMultiSpanProcessor threadpool should be reset after process forked

[Zhouhao12345]

[Environment]

python==3.7.1
opentelemetry-api==1.7.1
opentelemetry-exporter-jaeger==1.7.1
opentelemetry-propagator-jaeger==1.7.1
opentelemetry-sdk==1.7.1

ConcurrentMultiSpanProcessor

[Issue]

ConcurrentMultiSpanProcessor setup thread executer to submit task for span exporter, but never reset it after process forked. This would block next span submit.

[expected behavior]

It should be reset after process fork, like
import os os.register_at_fork(after_in_child=lambda :reset_concurrent_span_processor())

[gregoiredx]

Let's give this a try #4862