Add optional rzCOBS encoding+framing

[Dirbaio]

Part of #492

Encoding is chosen with the encoding-raw, encoding-rzcobs features. If no encoding is chosen, rzcobs is used.

The used encoding is stored in the ELF symbol table, and the decoder automatically picks the correct version.

[japaric]

left some inline comments / questions. does probe-run need changes after this has landed? IOW, will landing this break (e.g. not compile / fail to decode) probe-run even if it uses the right decoder version?

EDIT: would sticking to the "raw" encoder / decoder for a bit not immediately break probe-run?

[Dirbaio]

will landing this break (e.g. not compile / fail to decode) probe-run even if it uses the right decoder version?

I think it'll still build because I've left Table::decode as-is. Since the default is rzcobs, it will fail to decode, though. It should switch to the StreamDecoder API.

EDIT: would sticking to the "raw" encoder / decoder for a bit not immediately break probe-run?

Yes, raw is identical to the previous behavior, just concatenating all frames with no framing. But since it'll need a release anyway to update to the new defmt-decoder (wire format has breaking changes) it might as well be updated to StreamDecoder from day 1, the required changes are tiny.

[japaric]

Yes, this is from the rzcobs crate. For encoding it's not quite right, for decoding we could use it as-is. I thought it was strange to have an in-tree impl for encoding and a crate for decoding, but can change it if you think it's better to use external crates.

In that case, it seems fine to "vendor" that code. This looks good to me.

bors r+

[bors]

Build succeeded:

• ci

[Urhengulas]

@Dirbaio Can you please fill the empty doc-comments on the methods of the encoder?

#539 (comment)

[Dirbaio]

Added missing docs in #568