[Code scan] Respect model precision without leaking torch default dtype

This issue was found by a Codex global scan of the repository at commit 19f9265efe8825a4383e5a13ce128361533b5ac6.

MACE accepts `precision` through config, but direct construction drops it through `**kwargs` and builds the MACE network from the ambient `torch.get_default_dtype()`:

https://github.com/deepmodeling/deepmd-gnn/blob/19f9265efe8825a4383e5a13ce128361533b5ac6/deepmd_gnn/mace.py#L270-L291
https://github.com/deepmodeling/deepmd-gnn/blob/19f9265efe8825a4383e5a13ce128361533b5ac6/deepmd_gnn/mace.py#L296-L315
https://github.com/deepmodeling/deepmd-gnn/blob/19f9265efe8825a4383e5a13ce128361533b5ac6/deepmd_gnn/mace.py#L338-L359

The `get_model()` config path sets the process-global default dtype and never restores it:

https://github.com/deepmodeling/deepmd-gnn/blob/19f9265efe8825a4383e5a13ce128361533b5ac6/deepmd_gnn/mace.py#L1823-L1831

I verified both effects locally:

```text
MaceModel(..., precision="float64") with global float32 builds atomic_energies as torch.float32
MaceModel.get_model(..., precision="float64") leaves torch.get_default_dtype() == torch.float64
```

NequIP has the complementary failure mode: it passes `model_dtype`, but does not temporarily raise PyTorch's default dtype before `model_from_config()`, so `precision="float64"` can fail when the caller's default dtype is still float32.

https://github.com/deepmodeling/deepmd-gnn/blob/19f9265efe8825a4383e5a13ce128361533b5ac6/deepmd_gnn/nequip.py#L258-L299

Local reproduction:

```text
ValueError: Overall default_dtype=float32, but model_dtype=float64 is a higher precision- change default_dtype to float64
```

Suggested fix: validate `precision` explicitly in both constructors, build the underlying MACE/NequIP networks under a temporary dtype context, store the effective precision in serialized params, and always restore the previous global default dtype.