[gpm]stale pid file

[Alessandro Rubini]

Please note that gpm is now being massively reworked, so some problem
does occasionally creep up.

> second, before claiming that gpm is already running, it ought to at
> least send a signal 0 to that pid to see if there even is such a pid...

This was one of the first things I did (1995, I think). Then I also
made the new gpm connect to the socket to see if that pid is really
gpm and not something else. This because running gpm at boot there
often was another process running with the same pid as gpm on the
previous boot (most often the same gpm process that complained);
sending signal 0 was not enough.