Filter based on additional metadata fields

[misko]

Extend metadata with additional fields present in the NPZ metadata file. Instead of just taking 'natoms' take everything that is present.
Allow user to specify a filter on the dataset in the format of

Take all indices that have absolute value of 'some_quantity' less then or equal to 5

subset_to:
  - op: abs_le
     metadata_key: some_quantity
     rhv: 5

Take all indices that a value of 'some_quantity' in [x,y,z]

subset_to:
  - op: in
     metadata_key: some_quantity
     rhv: [x,y,z]

[rayg1234]

why even have this class if we can just dump random things into it? might as well just make it a dict? not sure if this will break alot of things lol

[lbluque]

I think its better to keep it as a NamedTuple or a dataclass and fields that we decide are needed. This is a much more disciplined interface, so that we are not just dumping random things.

[misko]

We discussed this offline and agreed to follow up after this PR and figure out a better way to include Metadata inside of dataset object and make it fast enough to filter on.

[zulissimeta]

Is there a way to keep this info self-contained within the dataset files themselves? Every time we have another file that needs to be kept consistent.

This functionality is pretty close to what the ASE datasets already allow (filtering by natoms, or other fields in the object).

[codecov]

Codecov Report

Attention: Patch coverage is 81.25000% with 3 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/fairchem/core/datasets/base_dataset.py	81.25%	3 Missing ⚠️

Files with missing lines	Coverage Δ
src/fairchem/core/datasets/base_dataset.py	90.90% <81.25%> (-2.32%)	⬇️

[wood-b]

works for me