This is a bit of a stab in the dark for a flake sometimes seen on bots (but I can't repro locally) #20822. Detaching from a Flutter app currently seems to terminate it, so it may be the cause of the "app not found" when trying to stop the app subsequently.

This change means we stop the app from the main process first, before terminating the attach process.

* Unskip expression evaluation integration tests for non-Windows

* Add additional into to skip comments

* Add missing import

This test is sometimes flaky (#20822) because the two processes try to stop the same app. This fix changes to just gracefully terminate the attach process without explicitly trying to stop and then uses the original spawning process to stop the app.

I can't repro the flake locally to be certain, but I've verified only one stop command is sent now so it *should* be good.

Due to CPU contention we've seen these go really slow on Cirrus (see https://github.com/flutter/flutter/issues/19542#issuecomment-414265152), and there's also a chance our flakes are timeouts due to running slow rather than hanging.

All temporary directory start with `flutter_` and have their random component separated from the name by a period, as in `flutter_test_bundle.YFYQMY`.

I've tried to find some of the places where we didn't cleanly delete temporary directories, too. This greatly reduces, though it does not entirely eliminate, the directories we leave behind when running tests, especially `flutter_tools` tests.

While I was at it I standardized on `tempDir` as the variable name for temporary directories, since it was the most common, removing occurrences of `temp` and `tmp`, among others.

Also I factored out some common code that used to catch exceptions that happen on Windows, and made more places use that pattern.

* Upgrade everything except matcher.
* Roll matcher (and test)
* Adjust tests that depend on flutter:test directly to depend on a shim
* Require use of package:test shim and remove other references to package:test

* Skip tests instead of expecting failure, because expected failure = timeout which is slow

Some of these tests are known to fail and were coded to expect it; however because the failure is a timeout it's just dragging out test runs (shortening timeouts will make the tests flakier).

* Make it easier to find skipped tests

* Change SKIP to TODO

* Extract some of startApp into a reusable method

* Get basic attach --machine working

* Attach --machine tweaks

Move validation to validate method and create daemon early so we get the startup event before trying to get a connection.

* Bump daemon version so we know whether it's valid to flutter attach

* Tweak output text

* Swap package imports for relative

* Review tweaks (naming, formatting, typedefs)

* Separate arguments from process spawning

This will make calling attach easier

* Add a basic test for flutter attach --machine

* Fix crash if port unforward modifies the list of forwarded ports

* Add a no-op port forwarder for flutter-tester

* Switch to using BasicProject instead of our own inline code

* Fix expectation in test now we have a portForwarder

* Remove stale TODO (this is done)

* Tweak formatting

* Change some Completers to void to fix Dart 2 issues