[Feature] CatFrames used as inverse

[BY571]

Description

Adding the 'as_inverse' option to the CatFrames transform. With this transform we can use the CatFrames transform when adding data to the Replay Buffer. By that, the RB will consist of unfolded sequential samples.

Types of changes

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

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds core functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation (update in the documentation)
• Example (update in the folder of examples)

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

[vmoens]

We should also check what happens when we set as_inverse=True and append it to an environment. This should modify the observation in the input tensordict, which is pretty useless unless it is also part of the input keys.
So either we raise an exception when as_inverse=True and the transform has a parent or we make it possible but then we need to define what behaviour is expected and what transformation the observation_spec / input_spec should have. I'm more in favour of the first option (raising an exception if it has a parent).

[vmoens]

Almost there, can you add a simple check in test_transform_as_inverse where you create the transform and append it to an env, and check that the error is raised appropriately?

[vmoens]

LGTM (final)