batchspawner/batchspawner: .clear_state() if spawning fails

- The query_job_state change meant we don't call .poll() within the
  .start() method anymore.  .poll() would .clear_state() if the job
  was not found.
- So old state was: .clear_state() was called if the spawn failed.
- Add in .clear_state() explicitely when spawning fails.
- I think that if there are exceptions in unexpected places, then
  state will still not be cleared.  It's no worse than before, so we
  can leave it is-is for now.

Author: rkdarst

batchspawner/batchspawner.py

@@ -390,6 +390,7 @@ async def start(self):
             else:
                 self.log.warning('Job ' + self.job_id + ' neither pending nor running.\n' +
                     self.job_status)
+                self.clear_state()
                 raise RuntimeError('The Jupyter batch job has disappeared'
                         ' while pending in the queue or died immediately'
                         ' after starting.')