profiler: stop using runProfile in tests

[nsrip-dd]

In general, we should prefer testing the profiler with as close to the API
customers will use as possible. For one, it gives us more confidence that
customers will get the data they expect. Also, it's been a pain to work on the
profiler when internals-only changes break tests that happen to depend on the
internals of the profiler, such as TestRunProfile. This change starts getting
rid of uses of runProfile in particular.

This change gets rid of the CPU and goroutine sub-tests in TestRunProfile since
they really don't assert anything meaningful (other tests already check that
those profiles are present, and mocking their contents doesn't tell us
anything on top of that). For the goroutine wait profile sub-tests, we can at
least remove the dependency on run profile. This change moves those sub-tests
to their own test functions, TestGoroutineWait and TestGoroutinesWaitLimit. For
TestGoroutineWait, we are basically testing converting a debug goroutine
profile to a pprof, so the tests opts to use goroutineDebug2ToPprof directly
rather than hook into the profiler. For TestGoroutinesWaitLimit, we use
startTestProfiler, which is more end-to-end, and make assertions about the
profiles we would upload and the logs we would emit instead of hooking in to
the profiler.

TODO - tackle the "delta" test. We already test the delta implementation pretty
extensively, but it might be good to have some kind of more "end-to-end" test
for that. If we can remove the bits where we hook into the profiler there, I
think we can get rid of the testHooks stuff.

[pr-commenter]

Benchmarks

Benchmark execution time: 2025-06-03 19:03:28

Comparing candidate commit 93dc756 in PR branch push-vvpltltrvozv with baseline commit 3f3373d in branch main.

Found 0 performance improvements and 0 performance regressions! Performance is the same for 21 metrics, 0 unstable metrics.