StreamWorkManager should not terminate worker thread in case of failure

In case of exception during the processing of the Work, the processing should continue (skipping work in error) to have the same behavior as the default WorkManager.

For the Stream WorkManager this is not the case:

• when there is an exception during work.cleanup after a failure
• when there is an error when reading the work from the stream (serialization or corruption)
  When this happens the computation thread terminates which can result in blocking completely the processing in the Work pool.

The Work.cleanup invocation must be protected, since 10.10 the computation policy used to run the Work should be set to continueOnFailure=true.