Make backtrace and/or core dump available when a test crashes

If a test crashes (or it hangs and we kill it), we should attempt to provide some kind of backtrace when that is reasonably doable on the test platform.

We already generate a stub "this test crashed" test log when a test crashes, so we could put the backtrace into that log.

On Unix platforms it would be possible to collect and store the core file for the raw data, but I'm not sure how usable that will be in general (how portable are core dumps?)