Allow several timeouts before caching

Sometimes compilation/execution spuriously times out. Whenever that happens, the result gets cached. Users then rerun expecting the problem to resolve by itself (rightly so!), but it doesn't because the old cached timeout is simply displayed again.

A workaround would be to add a random space/newline somewhere in the code and then retry.

This looks really bad, especially to beginners. And it happens frequently enough to give a bad impression and get annoying.

I propose caching the result of a run in these two cases (rather than always, unconditionally):

1. timeout didn't happen
2. same code timed out 5 times

What do you think? If you like this idea, I could write a PR.