[Core] Enable CUDA graphs for DP + All2All kernels

[varun-sundar-rabindranath]

Enable CUDA Graphs for DP + All2All kernels.

Fixes:

1. The input buffers to the quant_method aren't captured properly when using CUDAGraphs + torch.compile. This PR introduces a staging area where the hidden_states and router_logits are copied into and it is this tensor that gets passed into quant_method.
2. It is important that all DP ranks invoke the same number of dispatch and combine kernels. The kernels need to synchronize between DP ranks. When this requirement isn't respected, it manifests as a deadlock. To this effect, introduce a get_dp_padding method in gpu_model_runner.py.

Tests: Verified correctness using lm_eval locally on 4xH100.

[github-actions[bot]]

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

🚀

[varun-sundar-rabindranath]

@bnellnm @youkaichao @tlrmchlsmth PTAL! Thanks.