[connector/datadog] Fix leaking goroutines and enable goleak

[crobert-1]

Description:

goleak detected some leaking goroutines in this package. Some are due to issues we've already hit, some are actual leaks in the collector. The added Stop calls for the TraceWriter and StatsWriter were required to make sure we don't leak goroutines on shutdown.

Link to tracking Issue:
#30438

Testing:
goleak test is passing. Calling the Shutdown methods in the test now take quite a while due to how many goroutines are creating when the TracesAgent is created.

Documentation:
Help on the changelog message is appreciated, not sure what details would make sense to users and what wouldn't.

[crobert-1]

Looks like tests are failing due to timeouts from Shutdown taking too long.

The test output shows leaked goroutines, but I believe this is simply Shutdown not completing before it timed out.

goleak output:

goroutine 7 [chan receive]:
github.com/DataDog/datadog-agent/pkg/trace/writer.(*sender).loop(0xc000494fc0)
	/home/runner/go/pkg/mod/github.com/!data!dog/datadog-agent/pkg/trace@v0.50.2/writer/sender.go:170 +0x8c
created by github.com/DataDog/datadog-agent/pkg/trace/writer.newSender in goroutine 3
	/home/runner/go/pkg/mod/github.com/!data!dog/datadog-agent/pkg/trace@v0.50.2/writer/sender.go:163 +0x1a5

The senders are properly stopped by the datadog-agent TraceWriter.Stop() method.

It looks like datadogconnector.CreateTracesToMetrics creates 120 sender goroutines in its default configuration. NewTraceWriter creates another 16 goroutines (on my local machine). I'll test on my machine to see how long this ends up taking to complete (without a timeout).

[crobert-1]

Did some more investigation, this is blocked by #30487 as explained here.