embedded-io compatibility

[bugadani]

Resolves #652

I'm opening this PR early because I have a few open questions:

• embedded-io traits have an associated error type for IO errors. How should we incorporate this?
  • we can make EncodeError/DecodeError generic, and default to std::io::Error
  • we can introduce a custom io error type, map known std errors to it, and map embedded-io errors to EOF (ReadExact error) or Other.
• decode_from_embedded_io_read is a bit verbose. I think I'd try to create a single encode/decode function and abstract the std/embedded options with a bincode-provided marker trait. The implementation would be more complex, but the API would be simpler. What is your preference here?

Right now we can't have both std and embedded-io features active, because of the IO error question. I intend this as a temporary restriction and plan on lifting it if the resolution of that question allows.

[VictorKoenders]

Thanks for this!

embedded-io traits have an associated error type for IO errors.

I think make an EmbeddedIo { ... } variant, and put that behind the config flag. They're completely different errors from std's Io errors. In hindsight we maybe should've named the existing variant StdIo.

The error type is marked as #[non_exhaustive] so people cannot match on all variants, so I believe it is fine to add variants based on features being active

decode_from_embedded_io_read is a bit verbose.

What about bincode::embedded_io::decode_from_read or bincode::embedded_io::decode?

[VictorKoenders]

Also please add a line to the features table in src/lib.rs

[bugadani]

Current implementation ignores the actual embedded-io error for simplicity. Trying to make Decode/EncodeError generic is at this point not practical.

I have also fixed a few typos I found.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.