Avoid overlapping jobstore disk writes.

Given that a followup write could be scheduled while one is already
going on, if the first one took too long, then the send write could
start and overlap with the first write. This change prevents that
and makes sure there's only one write happening at a time.

Bug: N/A
Test: atest CtsJobSchedulerTestCases
Test: atest FrameworksServicesTests:JobStoreTest
Change-Id: Ie77510508795e5395a24a40266178283d0c525cb