Be flexible with Zarr 3 index codecs

[frcroth]

Currently, only bytes and crc32c are supported. Other codecs will have an impact on the compressed size of the index, which needs to be calculated based on the codecs present.

Raised in #7553

I guess this works for most datasets, but not for all. Although, it doesn't make a lot of sense, it is spec-compliant to have multiple crc32c checksums. In the future, there might be other codecs that change the size of the shard index.

A more correct way would be to compute the size of the index based on the entire index_codecs pipeline. zarrita uses for that the compute_encoded_size methods.