Adds a multi property example

[bananenpampe]

Adds an example for multi property prediction, for the prediction of chemical shieldings and individual terms from the reference DFT calculations taht make up the isotropic target. Its an interesting example, because one of the properties is also a structure wise property (a global shape contribution).

This was intended to be a minimal working example to identify what breaks in #503.
It works with SOAP-BPNN, I keep trying to find out why it doesnt work with nanoPET

Contributor (creator of pull-request) checklist

• Issue referenced (for PRs that solve an issue)?

---

📚 Documentation preview 📚: https://metatrain--578.org.readthedocs.build/en/578/

[Copilot]

Pull Request Overview

This PR adds a minimal working example for multi-property prediction for chemical shieldings, targeting both local and global (structure‐wise) properties. The key changes include:

• New YAML configuration files for training (options.yaml) and evaluation (eval.yaml).
• Specification of multiple target properties with distinct settings for per-atom and global predictions.

Reviewed Changes

Copilot reviewed 5 out of 9 changed files in this pull request and generated 2 comments.

File	Description
examples/multi_property/options.yaml	Adds training, validation, and test set configurations for multi-property prediction using SOAP-BPNN.
examples/multi_property/eval.yaml	Provides evaluation configuration mirroring the training targets.

Files not reviewed (4)

• examples/multi_property/README.rst: Language not supported
• examples/multi_property/test.xyz: Language not supported
• examples/multi_property/train.xyz: Language not supported
• examples/multi_property/val.xyz: Language not supported