Fix load_weights with strict=False to filter extra weights before update

[gmin7]

Proposed changes

When loading weights from a checkpoint that contains more layers than the model (e.g., loading a full
model's safetensors into a model instantiated with num_hidden_layers=1), load_weights(..., strict=False)
raises an IndexError: list index out of range. This happens because indexed keys like layers.1.weight
pass through tree_unflatten and Module.update tries to index into the model's layers list at positions
that don't exist.

This restores the filtering of weight keys to only those present in the model's parameters when
strict=False, so extra weights are silently dropped before reaching update.

Checklist

Put an x in the boxes that apply.

• I have read the CONTRIBUTING document
• I have run pre-commit run --all-files to format my code / installed pre-commit prior to committing changes
• I have added tests that prove my fix is effective or that my feature works
• I have updated the necessary documentation (if needed)

[angeloskath]

Thanks for the catch and fix!