Streaming datasets V2

[knighton]

Formats: MDS, JSON, XSV (CSV, TSV)
Sample sizes moved to shards
V2 small JSON index format (faster startup time at scale)
Sample fields now have types instead of having to manually provide decoders
Prefetch factor
Fast process pool Dataset.download()
Dataset supports random access of samples, lazily loading their shards
Offsets live on disk instead of memory for scalability (two seeks instead of one for sample access)
Various compression algorithms
Various hash/checksum algorithms on shards (TODO: on index)
Datasets: CIFAR10, ImageNet (TODO: more)

[hanlint]

Before merging, could you setup:

• github actions for pytest, yahp, pyright, etc. ( can copy from https://github.com/mosaicml/mcontrib )
• move your testing into a tests/ folder and verify that pytest works

This would at minimum auto-format your code into your yahp standard to reduce diff noise / conflicts in the future.

[hanlint]

Didn't review for correctness, but some suggestions from API perspective.

[hanlint]

yolo