[service-bus] Making BatchingReceiver properly cleanup after an abort (Azure#13073)

Some more fixes related to the prior PRs around making BatchingReceiver behavior consistent:

* When we are closed (without error) or if we're in receiveAndDelete mode, we use the setTimeout() to ensure that we run after all pending message callbacks have fired
* We always cleanup our handlers before resolve/reject and it's made more clear if we're resolving immediately or resolving after pending message callbacks.

I also did some small refactors to make unit testing a bit easier, which should help make it simpler to validate this code path in the future.

Fixes Azure#12922

Author: richardpark-msft

sdk/servicebus/service-bus/CHANGELOG.md

@@ -5,6 +5,8 @@
 - Fix the `isNode` check to allow the package to be usable in Electron. [Bug 12983](https://github.com/Azure/azure-sdk-for-js/issues/12983)
 - Fix issue where receiveMessages might return fewer messages than were received, causing them to be potentially locked or lost.
   [PR 12772](https://github.com/Azure/azure-sdk-for-js/pull/12772)
+  [PR 12908](https://github.com/Azure/azure-sdk-for-js/pull/12908)
+  [PR 13073](https://github.com/Azure/azure-sdk-for-js/pull/13073)
 - Updates documentation for `ServiceBusMessage` to call out that the `body` field
   must be converted to a byte array or `Buffer` when cross-language
   compatibility while receiving events is required.