Correctly handling failed execlp calls

[Vojtěch Štěpančík]

This is probably a very minor edge-case, but when xbindkeys fails to exec `sh` in a forked process, it just jumps back into the event loop, so with every failed exec there's now a new xbindkeys process, which eventually leads to a failed assert in xcb and a crash.

The correct behavior is to kill the forked process when exec returns, because it only returns when it fails. I'm attaching a patch.

Vojta

exec_fail.patch
Description: exec_fail.patch