workflows should "fail" if an HPC job fails

[tiborsimko]

Currently, HPC compute backend jobs fail, but the workflow is reported as being "finished". Should rather be reported as "failed". for consistency.

Example: reanahub/reana-demo-worldpopulation#38

$ reana-client status -w hello-cwl-hpc            
NAME            RUN_NUMBER   CREATED               STARTED               ENDED                 STATUS     PROGRESS
hello-cwl-hpc   1            2021-02-18T17:52:35   2021-02-18T17:52:35   2021-02-18T17:53:42   finished   0/1   

$ reana-client logs -w hello-cwl-hpc | grep Status
==> Status: failed

Note: Serial and Yadage report well:

$ reana-client list | grep hpc
world-yad-hpc          1            2021-02-21T10:13:23   2021-02-21T10:13:24   2021-02-21T10:19:33   failed  
hello-yad-hpc          1            2021-02-18T14:20:26   2021-02-18T14:20:27   2021-02-18T14:21:17   failed  
world-ser-hpc          1            2021-02-21T09:45:32   2021-02-21T09:45:33   2021-02-21T09:51:18   failed  
hello-ser-hpc          1            2021-02-18T17:38:13   2021-02-18T17:38:14   2021-02-18T17:39:18   failed