[wptrunner] Check we actually manage to close the window
After
bef892c,
Safari had a number of tests change behaviour:
https://wpt.fyi/results/?diff&filter=ADC&run_id=5195649314455552&run_id=5132128627195904
Ultimately, this is because we practically checked windows closed
before, because the "Hope the first one here is the test window"
comment turned out to almost always be untrue in Safari.
This actually leaves us somewhere better than that, as it makes the
check explicit, and as long as close_after_done is True (which it
typically is), it makes the test that left us in the broken state come
back as ERROR, rather than the following test.