rust-amqp is too prone to panics

[lucab]

There are many panic-points in rust-amqp and most of them are undocumented. Applications can easily blow up on those, so there is work in progress to remove those in favor of traditional error signaling.

Culprits observed so far:

• [x] panic!()
• [x] .expect()
• [ ] .unwrap()

[Antti]

Removed most of the unwrap calls, there are still few left in basic, session and examples.

[onelson]

I'm mitigating by wrapping each place I call amqp in a thread spawn so that I can retry on error. Is there a better way to manage this?