This mirrors what we did in jmp-fwdcalls commit 5901c10, but also adds
some more newlines to make it look nicer (which jmp-fwdcalls had
already).

Note that both mpx-catapult and sgx-catapult are being updated here;
aside from the new log line printing, there are no functional changes.

The "catapult_num-N" keys have always stored the JID associated with N
so it makes more sense to call this "catapult_jid-N" - this is really
a "JID key", not a "number key" (hence the accompanying variable name
change).  Furthermore, we will be adding another number-indexed table
to the schema soon ("catapult_fwd-N"), which will eventually be used
for forwarding calls, but is initially mostly a placeholder in the DB,
used only for sending verification codes via phone from jmp-register.
It makes much less sense to have "catapult_num-N" in this situation.

This is for compatibility with jmp-register commit f7ee1fe (whose code
is not yet active), and for general simplicity and correctness.  We
can now use SETNX to mitigate races rather than set-then-check-length.

Thanks to singpolyma for the suggestion; use more Ruby-isms as I can.

This change was suggested by singpolyma; looks more correct to me too.

* include new copyright line, as the file has an additional author now
* split up the `require` lines a bit as we're not alphabetical anymore
* move/rename the HTTP listen port in the command-line parameters
* revert ARGV change since others missed in 26d6c4f; reordering anyway
* add comment to #write so I remember why it was added to begin with

Demonstrates how to integrate Goliath into your EM reactor alongside
Blather.  And also how to send an XMPP stanza in response to a webhook.

We used to catch all iq stanzas and do our own checks to see if they
matched the element name and namespace we were looking for.  Something
like this:

```
	query_node = i.children.find { |v| v.element_name == "query" }
	if query_node.namespace.href ==
		'http://jabber.org/protocol/disco#items'
		...
```

But singpolyma pointed out that this isn't necessary.  So now we use
Blather-native message filtering, and things are much prettier.  And,
more importantly, they are also much more correct.

This diff is fairly small because the largest blocks remain the same
(with the same indentation).  Smaller blocks were moved, de-indented.