LocalJobProvider should return an accurate start-time

[bfcrampton]

A job's start-time for the Google provider reflects "The time at which the job began to run" docs. This field is only populated once the vm has been created on Google Cloud and the job is in a true sense running (and it can now be cancelled). The Local provider, however, returns the start-time as the time at which the docker command was executed, not once the docker container is created. It would be great if this could be consistent, and the start-time was updated from within the container right before running the script.

In practice this prevents things writing tests that start a local job, poll dstat until the start-time is present, and then abort the job without a workaround like:

job = self.start_job('sleep 30')
time.sleep(10)
wait_for_job_status(job, ApiStatus.RUNNING)
abort_job(job)
wait_for_job_status(job, ApiStatus.ABORTED)

[bfcrampton]

Given recent updates to expose events in the local provider, I believe this can be fixed relatively easily by looking at the timestamp of the first event in events.txt here instead of just using the create-time.