Fix parallel notebook

[ChristopherMayes]

Summary

This PR fixes two Pydantic warnings that appeared when running Xopt on Python 3.14, and adds the missing gest-api dependency to the conda environment.

Changes

• Replaced the deprecated from pydantic.v1.json import custom_pydantic_encoder import in xopt/pydantic.py with a small, self-contained encoder. This removes the repeated "Core Pydantic V1 functionality isn't compatible with Python 3.14 or greater" warning while preserving the original encoding behavior.
• Declared _loaded_population and _toolbox as PrivateAttr in xopt/generators/ga/cnsga.py. These attributes were previously set in __init__ without being declared, which caused Pydantic to emit PydanticSerializationUnexpectedValue warnings during serialization.
• Added gest-api to environment.yml so the dependency is installed with the conda environment.

Testing

• Restarted the notebook kernel and re-ran docs/examples/basic/xopt_parallel.ipynb. The previously affected cells now run cleanly with no warnings.
• Verified in a fresh interpreter that importing xopt.pydantic no longer raises the Pydantic v1 warning and that the custom encoder correctly serializes NumPy values.

[codecov]

Codecov Report

❌ Patch coverage is 36.36364% with 7 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
xopt/pydantic.py	12.50%	7 Missing ⚠️

📢 Thoughts on this report? Let us know!

[electronsandstuff]

Looks ok, we should eventually modernize how datatypes like np.ndarray and torch.Tensor are serialized instead of integrating pydantic v1 things here, but I think it gets it working for the meantime. I will put in an issue about it to remember.

Also, run pre-commit run --all first.

[electronsandstuff]

Use more modern syntax. Also suggest for other fields in this file.

Suggested change

	_loaded_population: Optional[pd.DataFrame] = PrivateAttr(None)
	_loaded_population: pd.DataFrame | None = PrivateAttr(None)