Permit conversion into and creation from parts.

[mxinden]

Copy of matthunz/futures-codec#54 by @romanb.

This PR generalises Framed::release() et al to Framed::into_parts(), providing corresponding Framed::from_parts. This is pretty much analogous to tokio_util::codec::FramedParts. There are mainly two use-cases:

1. Resuing framed buffers with a new codec ([tokio-rs/tokio#717](https://github.com/tokio-rs/tokio/issues/717)).

2. Protocols that use a particular `Framed` only e.g. during a handshake phase after which they want to drop down to direct use of the underlying I/O stream. Since the `Framed` may have read and buffered data beyond the handshake frames, it is necessary to drain the remaining read buffer first when doing this. While one can already obtain the read buffer from `Framed::read_buffer()`, using it to capture the buffer before calling `Framed::into_inner()` requires to unnecessarily clone the `BytesMut`.

I also tweaked the documentation a bit to make it clearer that the into_inner() methods may discard buffered data. Since this PR removes / renames Framed::release() et al to into_parts, the API changes are not backward-compatible.

@romanb do you consent to your work being used in asynchronous-codec?

[romanb]

@romanb do you consent to your work being used in asynchronous-codec?

Sure!