Improve error handling when running tests

[adam-fowler]

Instead of just reporting there was an error, actually review the error and respond. Also moved error checking to the correct catch block

• If process return 1 and was not killed. Assume test failed and ignore
• If process was killed report in console
• If process signal is SIGILL then it crashed, report in console and attach last line of stderr as error for currently running test
• Unrecognised errors are reported to the console

Also execFileStreamOutput no longer pointlessly wraps the error in {}