Fix loading KV quantization scale; Enable modelopt kv cache

[yundai424]

Motivation

Currently kv cache quantization is not usable because 1) we never pass layer.k/v_scale to attention backend, 2) we never load the precomputed kv scale.

TODO: kv cache scale is currently only supported in flashinfer backend.

Modifications

1. Add logics to load precomputed kv scale if present in modelopt
2. pass RadixAttention's k/v scale to flashinfer backend

Checklist

• Format your code according to the Code Formatting with Pre-Commit.
• Add unit tests as outlined in the Running Unit Tests.
• Update documentation / docstrings / example tutorials as needed, according to Writing Documentation.
• Provide throughput / latency benchmark results and accuracy evaluation results as needed, according to Benchmark and Profiling and Accuracy Results.
• For reviewers: If you haven't made any contributions to this PR and are only assisting with merging the main branch, please remove yourself as a co-author when merging the PR.
• Please feel free to join our Slack channel at https://slack.sglang.ai to discuss your PR.