add Ipld de/encoding for CACAOs

[chunningham]

Implements Encode<DagCborCodec> and Decode<DagCborCodec> for Payload and CACAO, to support encoding cacaos as IPLD blocks. Also adds a test case taken from the ceramic cacao-ts tests (currently fails). Closes #4

[clehner]

Looks good... I confirm the test fails, with the following:

thread 'tests::test_ipld' panicked at 'called `Result::unwrap()` on an `Err` value: Missing key `aud` for decoding `cacao::payload_ipld::TmpPayload`.', src/lib.rs:308:10

... which is strange because the linked test in Ceramic's implementation includes an "aud" payload value. But I didn't try to check that by decoding the binary data.

Another possible test case may be the one in the CACAO CAIP: https://github.com/ChainAgnostic/CAIPs/blob/8fdb5bfd1bdf15c9daf8aacfbcc423533764dfe9/CAIPs/caip-draft_cacao.md#test-cases

[clehner]

Test decoding from CBOR passes in #12. Round-trip encode/decode succeeds in #13 (However, the serialization is different from the bytes that the Ceramic test vector uses, as some properties are in a different order; therefore different CIDs result...).

I think maybe the dependency on DagCbor in the types could be reduced by using Codec Decode + Encode instead. But to use CIDs it it looks like we have to pick a serialization. CAIP-74 suggests CBOR, and that is what Ceramic's test case uses, so I think we'll be using DagCbor anyway, although we could allow use of other serializations for the CIDs.

[chunningham]

It would be ideal to have Encode<C> + Decode<C> where C: Codec without hard dependancy on one, but yeah there are optionality challenges with transliteration in IPLD (only the 3 DagX codecs would preserve any meaningful structure, so de/encode can't be so generic without a different StructuredCodec marker trait maybe). w.r.t. CIDs yeah we have to choose our default for many purposes, but I agree we should allow for other block producers to choose an alternative. CBOR was really used for convenience w.r.t. available tools + seems to be the go-to for most projects.

By rearranging the TmpPayload fields in their deterministic order, the decoding test is fixed 👍