Variable Layer Dropout

[mostafaelhoushi]

What does this PR do? Please describe:
It enables specifying a different dropout rate for each layer.
Basically you can set a layer drop probabilities using:

# all layers will have drop_p set to 0.2 (same as before)
>>> model.layers.drop_p = 0.2

# all layers will have drop_p set to 0.2
>>> model.layers.drop_p = get_values(scale_period=len(model.layers.drop_p), max_val=0.2) 

# every other layer to have dropout of 0.2 This implements the [LayerDrop](https://paperswithcode.com/paper/reducing-transformer-depth-on-demand-with-1) paper
>>> model.layers.drop_p = get_values(scale_period=len(model.layers.drop_p), max_val=0.2, slice_str="1::2")

# dropout increasing linearly from 0 to 0.2 across layers. This implements the [Progressive Layer Dropout](https://arxiv.org/abs/2010.13369) paper.
>>> model.layers.drop_p = get_values(scale_type="linear", scale_period=len(model.layers.drop_p), max_val=0.2)

# dropout increasing exponentially from 0 to 0.2 across layers. This implements the training recipe of [LayerSkip](https://arxiv.org/abs/2404.16710) paper.
>>> model.layers.drop_p = get_values(scale_type="exp", scale_period=len(model.layers.drop_p), max_val=0.2)

!closes facebookresearch#640

Does your PR introduce any breaking changes? If yes, please list them:
No. I ensured that setting a single value for drop_p works by overriding setters and getters.

Check list:

• Was the content of this PR discussed and approved via a GitHub issue? (no need for typos or documentation improvements)
• Did you read the contributor guideline?
• Did you make sure that your PR does only one thing instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests?
• Did you verify new and existing tests pass locally with your changes?
• Did you update the CHANGELOG? (no need for typos, documentation, or minor internal changes)