forEach tests using a stopwatch are flaky

[natebosch]

For example:

pool/test/pool_test.dart

Lines 558 to 561 in 90eab06

	expect((watch.elapsed ~/ itemCount).inMicroseconds,
	lessThan(expectedElapsed / (poolSize - otherTaskCount)),
	reason: 'Average time per task should be '
	'proportionate to the available pool resources.');

Example failure:

Expected: a value less than <476.1904761904762>...
Expected: a value less than <476.1904761904762>
  Actual: <542>
   Which: is not a value less than <476.1904761904762>
Average time per task should be proportionate to the available pool resources.
package:test_api                                             expect
google3:///third_party/dart/pool/test/pool_test.dart 558:15  main.<fn>.<fn>.<fn>
===== asynchronous gap ===========================
dart:async                                                   _AsyncAwaitCompleter.completeError
google3:///third_party/dart/pool/test/pool_test.dart         main.<fn>.<fn>.<fn>
===== asynchronous gap ===========================
dart:async                                                   _asyncThenWrapperHelper
google3:///third_party/dart/pool/test/pool_test.dart         main.<fn>.<fn>.<fn>

My guess is the host happened to run a bit slow, or we hit GC, or something like that. We either need to split these out into a separate performance test file that we can avoid running most of the time, or we need to make it less sensitive to whatever might cause it to spike.