Failing Dataflow job should explicitly state reason to be missing container.

[https://builds.apache.org/job/beam_PreCommit_Python_Phrase/903/](https://builds.apache.org/job/beam_PreCommit_Python_Phrase/903/)

Currenlty job timeouts or states that no values were returned. Job should fail with explicit reason of "container missing".

Current relevant log is description of job containing:

"experiments": [ "use_staged_dataflow_worker_jar", "use_fastavro", "runner_harness_container_image=gcr.io/cloud-dataflow/v1beta3/harness:2.17.0", "use_multiple_sdk_containers" ],

Following container image path shows us that container is missing.

Imported from Jira [BEAM-8506](https://issues.apache.org/jira/browse/BEAM-8506). Original Jira may contain additional context.
Reported by: Ardagan.