OtlpGrpcSpanExporter's shutdown can hang forever if the Channel is already shut down

[andimiller]

Describe the bug
OtlpGrpcSpanExporter's shutdown can hang forever if the underlying Channel is already shut down

Steps to reproduce

1. Create an OtlpGrpcSpanExporter
2. Create a BatchSpanProcessor and feed in the exporter
3. Create an SdkTracer and feed in the processor
4. Create an OpenTelemetrySdk and set the tracer
5. Report some spans (optional)
6. Call shutdown() on the SdkTracer and wait for the callback
7. Call shutdown() on the BatchSpanProcessor and wait for the callback
8. Call shutdown() on the OtlpGrpcSpanExporter and wait for the callback
9. Observe that the callback never completes

What did you expect to see?
I expected the callback from OtlpGrpcSpanExporter's shutdown to complete

What did you see instead?
The program hung indefinitely

What version and what artifacts are you using?

libraryDependencies ++= Seq(
  "io.opentelemetry"  % "opentelemetry-exporter-otlp" % "1.2.0",
  "io.grpc" % "grpc-okhttp" % "1.38.0",
   // "io.grpc"           % "grpc-netty"                  % "1.38.0" // I also tried this
  "io.opentelemetry" % "opentelemetry-api" % "1.2.0",
  "io.opentelemetry" % "opentelemetry-sdk" % "1.2.0",
)

Environment
Compiler: Scala 2.13.4 and AdoptOpenJDK 11.0.3
OS: NixOS 20.09
Runtime: same JDK as above
Runtime OS: same as above

Additional context
I will follow up with a java example that demonstrates this

[andimiller]

https://github.com/andimiller/opentelemetry-shutdown-bug here's a repo that reproduces the bug in java, it's using sbt because I can't remember how to use maven