[Refactor] make to_tensordict() create a copy of the content

[nicolas-dufour]

Description

This PR aims to change tensordict.to_tensordict behaviour. The idea is to make a clone each time the method is called.

Motivation and Context

The motivation behind it is clone the tensordict when performing .to_tensordict.

• I have raised an issue to propose this change (required for new features and bug fixes)

Types of changes

What types of changes does your code introduce? Remove all that do not apply:

• New feature (non-breaking change which adds core functionality)

Checklist

Go over all the following points, and put an x in all the boxes that apply.
If you are unsure about any of these, don't hesitate to ask. We are here to help!

• I have read the CONTRIBUTION guide (required)

[vmoens]

Thanks @nicolas-dufour for this.
Here's how I would see this instead:

For TensorDictBase, return a tensordict like this

TensorDict({key: value.clone() if not isinstance(value, TensordictBase) else value.to_tensordict() for key, value in self.items()}, **metadata)

And metadata are all the values precomputed elsewhere EXCEPT memmap (it won't be memmap because we clone) shared mem (idem). We can leave the metatensor empty (since the metadata will change).

For savedtensordict we would need to do the same but load the tensordict first.

Hope that makes sense!

[vmoens]

LGTM! Let's address these two comments and make the tests pass, then we're good to go!

[vmoens]

LGTM! Let's address these two comments and make the tests pass, then we're good to go!

@nicolas-dufour also we should rename this PR
why not: "make to_tensordict() create a copy of the content"?

We should also update the docstrings