[exporter/azuremonitor] Fix memory leaks on shutdown

[crobert-1]

Description:

goleak was detecting a few leaking goroutines in this exporter. This change fixes one, and ignores the other as it's a result of the third party dependency. I've filed microsoft/ApplicationInsights-Go#70 to track the ApplicationInsights goroutine leak.

This change does the following:

1. Implements Shutdown for metric, trace, and log exporters. This Shutdown needs to call Close on the transportChannel, as documented here.
2. Copy trace exporter tests for metrics and logs. These are the tests added in factory_test.go, adding wider coverage. Without these we may have missed the need to add a Shutdown call to metrics and logs exporters.
3. Update the mock channel functionality which I manually generated using mockery.
4. Add goleak check.

Link to tracking Issue:
#30438

Testing:
All tests are passing.

[crobert-1]

Let me know if you disagree with my reasoning here, happy to discuss further.

This is not an error might be a bit misleading as technically we want the channel Close returns to send a message before the timeout is hit, but I think it captures the idea.

[crobert-1]

Tests are failing due to microsoft/ApplicationInsights-Go#60. I'll investigate if there's some way to account for this on our end.

In my mind, anything is better here than current functionality, which abruptly shuts down without any kind of safeguards. From my understanding, the worst case of the data race is simply dropping data that was meant to be sent, but that's what we're currently doing every time anyways, since we're not even calling Close. I think if there's any chance it's successful it's better than 0% success. I'll let others weigh in here though.

[crobert-1]

There's no way to workaround this at this point, I'm going to close for now. The final solution may have to be to ignore the leak instead of failing on a race condition, unfortunately.