[FEATURE] Effective drop path.

**Is your feature request related to a problem? Please describe.**

While current drop path implementation in TIMM doesn't save computation resources, implementing a **true** drop path that ignores unnecessary tokens will significantly speed up training when `path drop ratio` is high (e.g. 0.3 or 0.4).

**Describe the solution you'd like**
Reference to: https://github.com/facebookresearch/dinov2/blob/c3c2683a13cde94d4d99f523cf4170384b00c34c/dinov2/layers/block.py#L110

I already implemented a modified Block that utilizes this function and it gives me a huge performance improvement. I can add it to PR #1835 if it's a good idea.

Start by reading the referenced DINOv2 block implementation and compare it with the modified Block proposed for PR #1835. Verify that high path-drop ratios skip unnecessary token computation while preserving model behavior, then benchmark training to confirm the expected resource savings.

Written by the indexing model from the issue text.