[core] moe fp8 block quant tuning support

[divakar-amd]

This PR is required for tuning models with fp8 block quantization (e.g. DeepSeekV3).
The current implementation of fp8 tuning does not take into account the block quantization and hence, when tuning DSv3, it does not factor in the weight scales. This PR adds this support.

[github-actions]

👋 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.

🚀

[mgoin]

The changes look reasonable to me, thanks for the PR! Have you tried making a config and testing it?