[WIP] experimental GraniteSWAForCausalLM support

[yannicks1]

Please do not review/merge. This work is very much in progress.

Description

Branch that allows loading GraniteSWA models with and without weights for testing model compilation and per-liminary bench-marking.

Note: While this branch is currently very hacky, the final PR will be light weight.

What will go away:

• dummy weight loading code (moved out into this separate PR Feature: support --load-format dummy for random-init model loading #1013 as not specific to GraniteSWA)
• manual vLLM model registration: ModelRegistry.register_model in platform.py (will disappear when the model is registered in vLLM (see this vllm branch here, specific commit))
• separate gated model load with manual configuration _granite_swa_variant_from_hf (will disappear when we get actual checkpoints with model configs)

Dependencies

• fms branch
• hf branch
• [vllm branch (not actually used as long as we register the model manually)]
• special branches of lower Spyre stack...

Related Issues

#1013

Test Plan

Checklist

• I have read the contributing guidelines
• My code follows the project's code style (run bash format.sh)
• I have added tests for my changes (if applicable)
• I have updated the documentation (if applicable)
• My commits include a Signed-off-by: line (DCO compliance)

[github-actions]

👋 Hi! Thank you for contributing.
Just a reminder: Make sure that your code passes all the linting checks, otherwise your PR won't be able to be merged. To do so, run ./format.sh.
Now you are good to go 🚀.

We also recommend installing prek and configuring it to check your code before every local commit.