Support index_codecs in zarr sharding

[normanrz]

The Zarr sharding specification was recently changed to introduce index_codecs. This is the PR with the changes to the spec: zarr-developers/zarr-specs#253 Briefly, the index is now treated as a regular array that needs to be passed through a configurable codec pipeline. The crc32c checksum is turned into a bytes->bytes codec. A limitation is that only fixed-size codecs (ie. no compression) may be used for the index to ensure that the size/offset can be determined beforehand.

[normanrz]

I updated the demo dataset https://static.webknossos.org/data/zarr_v3/l4_sample

[normanrz]

Also there is a new zstd codec coming: https://github.com/zarr-developers/zarr-specs/pull/256/files

[normanrz]

Also, the endian codec will be renamed to bytes. Probably best to accept/process both names for the time being. The endian configuration in the new bytes codec is optional for 1-byte data types (uint8, int8, bool).

[frcroth]

I can implement the zstd codec now, but it is not really specified how the checksum is used, so it would leave it for now.