Optimizes getters of data inside asset classes

[Mayankm96]

Description

Previously the root positions was obtained by performing the following steps:

1. Call PhysX getters to get transforms and velocities
2. Concatenating the two to make a single state tensor
3. Indexing this tensor to get the position

For applications where we only want the positions, the above is an expensive operation as we call CUDA-MemCpy twice and then do concatenation which adds overhead.

This MR simplifies the above process by having separate lazy buffers for poses, velocities and states.

Fixes # (issue)

Type of change

• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

Screenshots

Please attach before and after screenshots of the change if applicable.

Checklist

• I have run the pre-commit checks with ./isaaclab.sh --format
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• I have updated the changelog and the corresponding version in the extension's config/extension.toml file
• I have added my name to the CONTRIBUTORS.md or my name already exists there

[jtigue-bdai]

I think its a good separation for the pose vs velocity buffers. The added lazy buffer layer in the update call is also a nice addition. Tests all pass (at least the ones I think are dependant) so I think its in good shape. Can you add the performance graphs to the PR description for reference when we look back?

[Mayankm96]

Will share my logs here tomorrow. It's on a different PC. Might have made sense to run OSMO benchmarking but that seems like quite a bit of an effort.