Redo ProgressLog handling

[jas88]

Thread safe ProgressLog handling - retains lock while inserting multiple rows, does it all on a background worker thread, then waits for it to complete at the end of a load job. Working apart from intermittent issue within LoggingDatabase_TestActuallyCreatingIt (on MySQL, oddly)

[JFriel]

Change makes sense to me, couple Q's but that's mostly for my own understanding

[JFriel]

@jas88 I took a look at the LoggingDatabase_TestActuallyCreatingIt issue. Was fairly inconsistent for me, could cause the test to fail on both MySQL and MSSQL ~40% of the time. It was more consistently failing when the system had been idle for a while.
Adding a 1000ms timeout seems to resolve the issue - ran the same testing with 100% success. Fix adds 4s (2s realistically as it doesn't run Oracle or postgres) to the test suite runtime, are we ok with that?

[jas88]

@JFriel One second would probably have been OK, but didn't fix the root issue properly - the log flush was being skipped after the first call to the close method, so logs would be written "fairly soon ish" after the call. Just switching the flush and repeat-call check around fixes it properly, so the log is always flushed before that method returns. (1s is probably long enough on the fast CI VM Github provides, but with production use on much less powerful VMs with much more contention, I'm betting we'd see multi-minute delays at least...)