[python] Initial work toward PyTorch data loaders

[bkmartinjr]

This PR is superceded by single-cell-data/TileDB-SOMA-ML#1 and is being closed. No further work will be done on this branch.

---

Notes to reviewers:

• APIs have changed from the CZI contribution. The demo notebook (tutorial_pytorch.ipynb) has been updated to run with all API modifications.
• This PR is for an initial drop we plan to use for pre-release feedback. More work is required prior to release
• This is set up as a separate Python package, with its own config/build/etc.

---

This PR contains a PyTorch iterable-style DataSet/DataPipe for use with SOMA Experiment. Initial code contributed by the Chan Zuckerberg single-cell team. Modifications to the original code include:

• stand-alone Python package tiledbsoma_ml
• significant refactoring to improve performance
• CI and other config for stand-alone Python package
• improved context/config handling - pass through all user-specified configuration allowing use with other storage solutions, etc.
• fix lint and typing issues identified by the CI pipeline in this repo
• update doctrings and copyright
• some minor enhancements to unit tests
• added torch.utils.data.IterableDataset alongside the existing torchdata.datapipes.iter.IterDataPipe, acknowledging that the torchdata.datapipes is deprecated and slated for removal
• add Python 3.8 support
• API refinement to better align with ExperimentAxisQuery
• modifications to support DDP (mult-GPU) training.

API change summary:

• constructor takes ExperimentAxisQuery, rather than than an Experiment
• reworked I/O and shuffle chunk size param for better API UX

Benchmarking results vs the cellxgene-census loader will be forthcoming shortly.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 89.79%. Comparing base (6f723d4) to head (124a510).

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #2823       +/-   ##
===========================================
+ Coverage   61.10%   89.79%   +28.69%     
===========================================
  Files          95       40       -55     
  Lines       12382     4096     -8286     
  Branches      781        0      -781     
===========================================
- Hits         7566     3678     -3888     
+ Misses       4625      418     -4207     
+ Partials      191        0      -191     

Flag	Coverage Δ
libtiledbsoma	?
python	89.79% <ø> (+0.14%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Components	Coverage Δ
python_api	89.79% <ø> (+0.14%)	⬆️
libtiledbsoma	∅ <ø> (∅)

[aaronwolen]

This is awesome. I'm excited to start testing it out. The docs are comprehensive and super clear. My suggested changes are largely minor typos and small improvements for clarity/consistency.

[bkmartinjr]

PR superceded by single-cell-data/TileDB-SOMA-ML#1