Add client-side support for block-wise observe

[SimenZhor]

Hi, and thanks for maintaining this crate!

I needed client side support for block-wise transfers for observe notifications, so I thought I'd see if you'd be interested in these changes upstream as well. I'm not communicating with a server that's using this crate, so I haven't touched the server side of the code at all. Let me know if you need anything changed!

• Set up synchronized transport before observe registration (to prevent losing first packet).
• Ensure that CoAP messages are populated with tokens.
  • "fixed" token for long-lived receiver in observe loop.
  • new token per block-wise datagram (all blocks in a datagram use the same token)
• Split out blockwise handling to more reusable components.
• Some formatting changes (by cargo fmt)
• Basic unit tests of new features

[coveralls]

coverage: 82.832% (+0.6%) from 82.27%
when pulling f63bf17 on SimenZhor:client-blockwise-observe
into 3850dc2 on Covertness:master.

[SimenZhor]

I've now created a separate function to generate random tokens instead of re-using gen_message_id.

I've also tried to investigate why my changes caused a drop in code coverage, and I think it's mostly because of the cargo fmt induced changes. Some calls that were previously on a single line have been broken into several lines, and the coverage tool doesn't seem to discover that all these lines are part of the same execution (?).

I did however notice some low hanging fruit while investigating, so I added a few very basic unit tests in the request.rs and server.rs test modules that should slightly improve coverage of those files.