OpenTelemetryDestroyer flushing

[jimkont]

following up from #35686 (comment)

in

quarkus/extensions/opentelemetry/runtime/src/main/java/io/quarkus/opentelemetry/OpenTelemetryDestroyer.java

Lines 17 to 18 in 09cd879

	openTelemetrySdk.getSdkTracerProvider().forceFlush();
	openTelemetrySdk.getSdkTracerProvider().shutdown();

Both calls return a CompletableResultCode, which is effectively a CompletableFuture. Should we use the .join(SECONDS, TimeUnit.SECONDS) in both calls to ensure the thread completes?
In the context of an HTTP server, with very few shutdowns, this should have a minimal effect, but in AWS Lambda calls, where the shutdowns are very frequent, it could possibly make a difference.

also related to #29448

[quarkus-bot[bot]]

/cc @brunobat (opentelemetry), @radcortez (opentelemetry)

[brunobat]

I think it makes sense. Let me hack something.

[jimkont]

fyi, when I try to inject a spanProcessor: SpanProcessor in the serving application and run spanProcessor.forceFlush().join(seconds, TimeUnit.SECONDS) inside the @PreDestroy hook, most of my errors go away.
if I put only spanProcessor.forceFlush() I still get errors.
using the forceFlush().join(...) I also do not need all the quarkus.otel.bsp.* configuration you mentioned here

[brunobat]

You should configure quarkus.otel.bsp.* I will not put 10s in the join's timeout... Most likely I will set a tight value.

[brunobat]

Created #36538.