Automatically retry enqueuing after connection reset

[borgand]

If DB connection is dropped for any reason, QC does reset the connection but does not retry enqueuing the job.

This PR adds retry-once logic to enqueue and enqueue_in (and through that to enqueue_at) so that when the initial execute fails with PGError, it is retried once (after execute's rescue has reset the connection).

Rationale: QC should give it's best to ensure the enqueue results in a enqueued job in DB, so if the DB connection has gone bad, resetting it is not enough and we should repeat the INSERT also. Leaving the retry logic to the app code is sub-optimal as then everybody need to implement the same code.
I added the retry only to enqueue and enqueue_in, not to execute, so that other SQL statements can be executed without automatically retrying on PGError.