Change process._tickCallback so that if a callback throws an error but
there are other callbacks after it, we indicate that
process._tickCallback needs to be ran again.

Currently, if a callback in process._tickCallback throws an error, and
that error is caught by an uncaughtException handler and
process.nextTick is never called again, then any other callbacks already
added to the nextTickQueue won't be called again.

Updated the next-tick-errors test to catch this scenario.

Note that "//" is still a special indicator for the hostname, and this does
not change the parsing of mailto: and other "slashless" url schemes.  It
does however remove some oddness in url.parse(req.url) which is the most
common use-case for the url.parse function.

Test running already has a timeout mechanism.

FIXME

- Add test case for pumping from unreadable stream.
- Document the sys.pump error handling behavior

If the function for a process.nextTick throws an error, then the
splice() never removes that function from the nextTickQueue array.  This
makes sure the functions that have been run in _tickCallback get removed
regardless of errors.

Also add a test for this.

OpenEmbedded doesn't like it when you look in /usr/lib

http://twitter.com/bascule/status/21626934378