Update to New Base Version

Seems to mostly just work.
Complete Uploads Once They Reach Size

The spec for Jingle claims either side can terminate the session, but suggests
that the receiver should, since it knows when it's received things.

The current code expects the sender to do it, and Conversations expects the
receiver to do it, leading to both sides waiting for the other to do something.

This change on its own is insufficient to actually fix this, but it's a big
step. We pull the size information we're given out of the session
initialization and store it in a cache, similar to how other session
information is currently stored.

Then, when the socks server detects a connection, it checks to see if there's a
size recorded for this connection. If there isn't, that's fine, it just does
what it currently does, which is hope the other side will close the socket when
it's done so the EOF will unstick things.
But if there is a size listed, then when it gets to full size it now returns.

In order for this to be fixed something will need to terminate the session
after the chunks are done being read. Eventually this should live in this repo,
but for now the code lives in Cheogram as part of the handler for the file
upload finishing.

Rather than try to fix that now, we'll do that later and just make sure we
return, so that code is able to run.

Implementation notes:
- hGet was changed to hGetSome because hGet will buffer until it gets the right
  amount of bytes, or EOF
	Since we're expecting the upload to hang when it's done, we need the code to
  unstick so we can count the new total and realize we're done. hGetSome is
  like a non-blocking hGet, but it it blocks on 0. So if there's no data, it
  will wait, but as soon as there's any data it will return right away.
- If there are multiple files in the same session that would be bad, but it
	seems like none of the supported clients do that. Other things already may
  not work in that case
Replace Non-Exhaustive Pattern with Case

I did this because in newer versions of GHC things like this require a
MonadFail instance, whereas explicitly handling it like this does not.

Now, we don't currently require those newer versions, but since I happened to
have a new one it was causing me trouble. And this new way works in both
versions, so I think it's safe to just throw it in here.
Current version in use by cheogram