Refactor anntorchdataset, add correct default dtypes

[martinkim0]

• Closes #xxxx (Replace xxxx with the GitHub issue number)
• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.
• Added type annotations to new arguments/methods/functions.
• Added an entry in the latest docs/release_notes/index.md file if fixing a bug or adding a new feature.
• If the changes are patches for a version, I have added the on-merge: backport to x.x.x label.

Summary

• Centralize and clean up code that validates getitem_tensors
• Now applies the correct default dtypes to different data registry keys. Before, everything was fetched as np.float32, but not continuous data is fetched as np.float32 and categorical data is fetched as np.int64. This leads to less dtype conversion downstream, e.g., an embedding module that does not have to convert indexes to Long.
• Cleaner way to force batched single observations
• Correctly applies sparse tensor conversion for backed sparse data now
• Adds more tests for AnnTorchDataset

[codecov]

Codecov Report

Patch coverage: 100.00% and project coverage change: +0.05% 🎉

Comparison is base (ed28ad7) 89.09% compared to head (5bc57c6) 89.14%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2250      +/-   ##
==========================================
+ Coverage   89.09%   89.14%   +0.05%     
==========================================
  Files         145      145              
  Lines       11817    11812       -5     
==========================================
+ Hits        10528    10530       +2     
+ Misses       1289     1282       -7     

Files Changed	Coverage Δ
scvi/data/_manager.py	98.52% <ø> (ø)
scvi/dataloaders/_ann_dataloader.py	93.33% <ø> (ø)
scvi/model/base/_training_mixin.py	95.45% <ø> (ø)
scvi/data/_anntorchdataset.py	95.83% <100.00%> (+5.47%)	⬆️
scvi/data/_utils.py	93.95% <100.00%> (+0.25%)	⬆️

... and 3 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.