Fail the run if the publish thread pool times out

[bentsherman]

The entire purpose of a pipeline run is to produce a particular set of results. If any of these results are missing for any reason, the run should be considered a failure. Otherwise the user has a false sense of success when in fact some results may be missing.

When all jobs are complete, the run waits for these two thread pools to finish:

nextflow/modules/nextflow/src/main/groovy/nextflow/Session.groovy

Lines 691 to 693 in 308d5a5

	// shutdown thread pools
	finalizePoolManager?.shutdown(aborted)
	publishPoolManager?.shutdown(aborted)

By default they will hard shutdown after 12 hours:

nextflow/modules/nextflow/src/main/groovy/nextflow/util/ThreadPoolHelper.groovy

Lines 47 to 51 in 308d5a5

	final delta = System.currentTimeMillis()-t0
	if( delta > max ) {
	log.warn(exitMsg)
	break
	}

Currently only a warning is logged, but the run is still considered successful as long as all jobs completed successfully (or were ignored).

Instead, the run should fail in this case. As a compromise, we could fail the run only if workflow.output.ignoreErrors is false, so that the user can keep the existing behavior if they want. But the default behavior should be to not declare success until all results have been published.