Improvents in MTR crash or deadlock reporting

Make MTR kill server with SIGABRT when the server startup timeout is exceeded

Now when the server does not start in reasonable time, the only diagnostics that we get from buildbot is

Failed to start mysqld.3

and the server error log. When server hangs on startup, we would want to have a stack trace.

Check if it's possible to kill the server with SIGABRT when the result is 'Failed to start'. If the server is already down, it won't hurt, but if it hangs, it can give more information.

Make MTR kill server with SIGABRT when the server shutdown timeout is exceeded and print the stack trace

When server shutdown is exceeded, MTR silently SIGKILLs the server and proceeds. It hides important failures on server shutdown. Instead, it should report a hang and get a stack trace. On the other hand, it should be differentiated from intentional server killing, when the shutdown timeout is initially zero – in that case, it should be SIGKILL-ed without a stack trace.