If we don't find it, we create a new one, and if we do find it, we
don't use it.

We're asserting now, and we would have already blown up from
dereferencing earlier.

switch_event_expand_headers should only return null on a malloc
failure, so we might as well just assert here.

The node could conceivably disappear before we grab the globals mutex
and we'd get back a null pointer.

This would have briefly prevented setting outbound_per_cycle from the
mod_fifo config file.

We already decided to goto end if node is null, so it cannot be null
here.

This would have briefly caused the importance value in the fifo config
file to be ignored.

FS-6513 --resolve This function was assuming the app was being run with execute_on_media so it did not properly check for established media.  Now it will use the scheduler method automatically when this state is detected.

Nike edition of makefile for quickly building debian7 or centos 6.5 freeswitch takes care of all deps.

Ask me sometime how I was "reminded" of this.

In this branch we made a pass through the code commenting on and
documenting the most important bits, and cleaning up and refactoring
where possible while maintaining existing behavior.

Thanks to Anthony, Mike, and Brian for explaining some of the history
of and rationale for what's in tree.

We already checked the number of usable rows, and nothing has changed.

We were simply discarding unused the first time we retrieved here.