usleep from coreutils != usleep from redhat

[Jasper Huijsmans]

Ok, 

usleep from coreutils sleeps for a number of _milliseconds_ instead of
microseconds. That sounds too silly to be true, but it is (imagine
writing a utility that basically just calls 'usleep(arg * 1000)', and
still calling it usleep ...).

So, what should I do? Remove it from coreutils (it's a special patch,
anyway) and put back the redhat one, changing the patch, or update the
scripts?

We can't have two usleep binaries doing different things.

	Jasper