feat: add reload argument to *Job.done() functions

[tswast]

This enables checking the job status without making an API call.

It also fixes an inconsistency in QueryJob, where a job can be
reported as "done" without having the results of a getQueryResults API
call.

Follow-up to #340

[shollyman]

Do the changes here indicate there's further changes for registering the done callback? My mental model has them independent, but I wanted to make sure.

[shollyman]

checking: What about DDL statements, where we have no results? Or would this already retain the empty row iterator?

[tswast]

DDL statements don't 404 for jobs.getQueryResults, do they? This is the full response object of jobs.getQueryResults, not just the rows.

[tswast]

Do the changes here indicate there's further changes for registering the done callback?

No callbacks here. Maybe when we implement the async client googleapis/google-cloud-python#15742

done does get called by result via PollingFuture._blocking_poll, which is why I can now remove the extra call to _get_query_results from QueryJob.result().

[tswast]

Do the changes here indicate there's further changes for registering the done callback?

Maybe you meant -- How do we get the needed arguments to get_query_results from result? I'm planning to do something a little bit hacky in https://github.com/googleapis/python-bigquery/pull/347/files -- which maybe is a bad idea, as it means you can't call result from two threads simultaneously. Hmmm...