Respect timeouts specified per query

[hogjosh]

Intended to fix #162, but I might be barking up the wrong tree! Maybe I can at least get the ball rolling...

[coveralls]

Coverage increased (+0.05%) to 66.104% when pulling 36e8734 on hogjosh:respect_query_timeouts into 452a42d on xerions:master.

[fishcakez]

This is good except the set charset query will have :infinity timeout during handshake and we don't want to allow that.

In postgrex this is handled by always using an :infinity timeout and then doing this: https://github.com/elixir-ecto/postgrex/blob/2c026cb29c4f8e62ce1b2fec46f444fac440ff4d/lib/postgrex/protocol.ex#L487-L510

[coveralls]

Coverage increased (+0.05%) to 66.211% when pulling 9a10be9 on hogjosh:respect_query_timeouts into 2972cb7 on xerions:master.

[hogjosh]

I fumbled through this. I'm happy to try to make additional changes.

If I'm understanding correctly, there's a recursive call to handshake_recv/2 and I wasn't sure if the timeout should be for the duration of the entire handshake or per call to handshake_recv/2

I also have no idea if the "shutdown" processing is what we had in mind.

[fishcakez]

I also have no idea if the "shutdown" processing is what we had in mind.

I think we can close socket, and let existing code handle socket closing error.