Move AsyncBatchingWorkQueue usage in telemetry to TelemetryLogging level

[ToddGrun]

Reapply "Move AsyncBatchingWorkQueue usage in telemetry to TelemetryLogging"

This reverts commit 4bbcd1d (which was a reversion of 21181a7).
Original PR which introduced most of these changes: #73287

Additionally:

1. Switch from AsyncBatchingWorkQueue to just a simple task based fire and forget mechanism. The first PR changed introduced the problem that we added work to the ABWQ immediately when the prior work completed. Thus the IAsynchronousOperationListener system would never see the queue as empty. As adding the work immediately avoided other issues, I wanted to keep that concept in this fixup. So, I instead moved away from ABWQ and just do a simple fire and forget to queue up the next collection.
2. Cleanup _findDocumentResults and _usedForkedSolutionCounter after their data has been flushed.

[ToddGrun]

Test insertion (as I broke insertions with the previous iteration of this) didn't hit the same issues as the previous PR was causing.

https://dev.azure.com/devdiv/DevDiv/_git/VS/pullrequest/551162

[CyrusNajmabadi]

What was different this time around,(will review when I get in the office)

[ToddGrun]

What was different this time around,(will review when I get in the office)

The first commit is strictly the revert. The second commit is the changes this time around. The issue was as you had originally thought, we were pushing an item into the ABWQ immediately after finishing the prior one (the code used to wait until a telemetry event was fired, but that didn't fit well with how things moved around). By pushing immediately into the ABWQ upon processing an item from it, the IAsynchronousOperationListener processing code would never complete.

[dibarbet]

consider wrapping with reportnonfatal (so we'll have non fatals if something goes wrong)