Implement distinct Flush types for Compress::compress vs Decompress::decompress

[chrisvittal]

This implements distinct flush types (FlushCompress and FlushDecompress) for Compress::compress and Decompress::decompress as in #79. It also removes the old Flush type.

To keep the code generic in read<R: BufRead, D: Ops> and Writer<W: Write, D: Ops>, this adds an internal Flush trait for abstracting over the common values of the new flush types. This new Flush trait is used as a bound for a new associated type (also called Flush) for Ops.

[alexcrichton]

Looks great to me! Just one minro comment about a little more flexibility around the enum, but otherwise I'd be ready to merge for the 1.0 release.

[alexcrichton]

Looks great to me! Just one minro comment about a little more flexibility around the enum, but otherwise I'd be ready to merge for the 1.0 release.

[alexcrichton]

While we're at it, both of these flush enums are intended to be extensible in the sense that client code shouldn't match on them exhaustively. To accomplish this mind adding a __Nonexhaustive variant tagged with #[doc(hidden)] to both enums? That way it should hopefully send a signal of "we may extend this in the future".

[alexcrichton]

👍 Thanks! I'll coordinate this with the 1.0 release so I won't merge this just yet, but rest assured I'll take it from here :)

[fstirlitz]

Why was this removed? The commit message doesn't explain.

[alexcrichton]

@fstirlitz I believe it's because it's not actually used in the backend, but do you have a use case for it?

[opilar]

@alexcrichton seems it's the last one already. Merge it and the crate is ready for evaluation!

[alexcrichton]

Indeed yeah thanks! I just need to allocate some time to go through the final steps of releasing 1.0 and then publish 1.0. I'll be sure to merge this before that though!

[alexcrichton]

Alright I'm doing some prep work so I've merged, thanks so much again @chrisvittal for taking this on!