std::io::Take should have into_inner() method

[lilyball]

It seems like a reasonably obvious thing to do and would be trivial to implement. The only alternative right now is to always use .by_ref() before calling .take(), and that's not always viable (e.g. because the io::Take needs to be kept around without borrowing the owner of the underlying reader).

An example usage would be implementing a message-based protocol on top of a stream protocol, where the messages either have fixed-length frames or have length indicators, I might want to call .take() to be able to work with a single message and then .into_inner() when I'm done to recover the underlying reader.

[huonw]

triage: I-nominated

A lot of library types have .into_inner methods, but I'm not sure we've ever explicitly decided that we should do it for any adaptor type like this (e.g. should the various iterator adaptors get .into_inners?), especially ones like this where &mut/.by_ref work for some/most(?) cases.

[alexcrichton]

triage: P-medium

@sfackler indicated that he was interested in making a PR for this and limiting it to the I/O adaptors for now.

[frewsxcv]

@sfackler Are you still working on this? Or can I take this?

[brson]

@frewsxcv you got it.