Add physical property prediction with PyTorch machinery

[mattwthompson]

Part of #148 - really just this block:

# Proposed fitlib.targets.thermo.predict body -- just passing msgpack trajectories in
def predict(
    data: datasets.Dataset,
    force_field: fitlib.TensorForceField,
    topologies: dict[str, fitlib.TensorTopology],
    simulations: tuple[
        dict[str, pathlib.Path], list[list[tuple[str, ...]]], pathlib.Path, pathlib.Path
    ],
    per_type_scales: dict[str, float],
) -> tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]:
    ... # returns predicted values, predicted stds, reference values, reference errors

Subsequent steps in this training loop will be handled elsewhere

[codecov-commenter]

Codecov Report

❌ Patch coverage is 10.12658% with 71 lines in your changes missing coverage. Please review.
✅ Project coverage is 82.40%. Comparing base (e6f019a) to head (dd4da08).

Files with missing lines	Patch %	Lines
tyff/targets/new_thermo.py	0.00%	55 Missing ⚠️
tyff/compute/workflow.py	20.00%	12 Missing ⚠️
tyff/mm/_ops.py	55.55%	4 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #190      +/-   ##
==========================================
- Coverage   83.44%   82.40%   -1.04%     
==========================================
  Files          58       59       +1     
  Lines        5195     5269      +74     
==========================================
+ Hits         4335     4342       +7     
- Misses        860      927      +67     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.