types: Investigate whether to use tmjson encoding on types shared between node + celestia-core

[renaynay]

We currently have a feature branch feature/api-breaks that will contain several breaking changes to the way some types are encoded in json, introduced by the Eiger team. The purpose of those changes was to ensure parity between the way types shared by node and celestia-core are encoded/decoded.

We currently have some fields (namely inside ExtendedHeader) that are tmjson encoded, so the proposal is just to tmjson the whole struct.

Originally, Preston from sovereign asked for parity encoding for those fields in #1476, referencing https://github.com/tendermint/tendermint/pull/7670/files which actually never ended up in celestia-core.

AFAIU, it seems the one thing that needs to be done to ensure parity between the encoding of types is to string marshal 64-bit integers which begs the question of whether we actually need to use tmjson at all, but rather just enforce string conversion for int64.

---