Add ScopedJobThreadPtr class to be used with JobThread - #4275

@borongc is it possible for you to help me review the part in this PR where all changes to Schedule() made during the scoped_ptr<> to unique_ptr<> migration is reverted?

@osagie98 @sideb0ard is it possible for you to take a look at this one?

@osagie98 @sideb0ard is it possible for you to take a look at this one?

looking now, re-reading bug history.

From reading Xiaoming and Jelle's comments, specifically https://b.corp.google.com/issues/372515171#comment11 this new ScopedJobThreadPtr has the behavior we require in the correct order, i.e. finish jobs, terminate thread, set pointer to null. However, as this PR and bug are still open, we appear to still have some lingering reservations.

Suggestion:

* Can we first add a new test to job_thread_test.cc 'ScheduleJobDuringShutdown', which specifically triggers the DCHECK listed in https://b.corp.google.com/issues/372515171#comment1
  This new test should FAIL with the current implementation.

* Then, secondly, run the new test with this PR, which should PASS. If it does pass, then we can be sure it's solved the issue as we currently understand it.

I've created PR for ScheduleJobDuringShutdown test: #4864

I've created PR for ScheduleJobDuringShutdown test: #4864

perfect, thank you!
Rather than merge #4864 which breaks the build, as expected:

Lets go ahead and add that test to this PR, which should now pass with these changes.

I've added ScheduleJobDuringShutdown test to this PR.

With this change, I think JobThread::ScheduleAndWait can be removed.

And we should also have another change to update tvOS code.

With this change, I think JobThread::ScheduleAndWait can be removed.

And we should also have another change to update tvOS code.

This PR will be followed by CL for the internal part.
But ScheduleAndWait() is still used in a few more places not related to this change (especially a lot in job_thread_test), should I remove it in this PR?

With this change, I think JobThread::ScheduleAndWait can be removed.
And we should also have another change to update tvOS code.

This PR will be followed by CL for the internal part. But ScheduleAndWait() is still used in a few more places not related to this change (especially a lot in job_thread_test), should I remove it in this PR?

I see, we can remove it in a follow up cl then.

The change breaks android on device testing. We need to fix the the failures before merging it. It seems that the failures are due to other issues.