[BugFix] Skip missing TensorDictParams submodule when loading state_dict with strict=False

[javierdejesusda]

Description

Loading a state_dict that has no entries under a TensorDictParams submodule
prefix crashes. In _load_from_state_dict, data = data.get(prefix) returns
None when the checkpoint contains nothing for that prefix, and the code then
calls self.data.load_state_dict(None) unconditionally, which raises. This is
the normal path for a strict=False load where a submodule is simply absent
from the checkpoint.

This PR guards the load so a missing submodule under strict=False becomes a
no-op, matching the nn.Module contract:

if data is not None:
    self.data.load_state_dict(data)

Motivation and Context

close #1472

A module holding TensorDictParams cannot be partially restored from a
checkpoint that omits those params, even with strict=False. nn.Module
treats missing keys as a no-op in that case, so TensorDictParams should too
instead of raising.

Types of changes

• Bug fix (non-breaking change which fixes an issue)

Checklist

• I have read the CONTRIBUTION guide (required)
• I have updated the tests accordingly (required for a bug fix or a new feature).
• My change requires a change to the documentation.
• I have updated the documentation accordingly.

[vmoens]

LGTM thanks!