sglang breaking changes

**CuteDSL BF16 GEMM on SM100 is on by default** when the heuristic allows it: [#30567](https://github.com/sgl-project/sglang/pull/30567)

**Breakable prefill CUDA graph is now on by default for DP attention**: [#31682](https://github.com/sgl-project/sglang/pull/31682)

**`sglang.jit_kernel` is retired into `sglang.kernels`**, completing RFC [#29630](https://github.com/sgl-project/sglang/issues/29630). Imports from the old module path must move: [#32072](https://github.com/sgl-project/sglang/pull/32072), [#31666](https://github.com/sgl-project/sglang/pull/31666), [#32015](https://github.com/sgl-project/sglang/pull/32015), [#32045](https://github.com/sgl-project/sglang/pull/32045)

**Chunked input-logprob processing is now on by default** to cap peak memory: [#31498](https://github.com/sgl-project/sglang/pull/31498)

**The experimental QServe (QoQ) W4A8 and FBGEMM FP8 quantization paths are removed** (per [#28543](https://github.com/sgl-project/sglang/issues/28543)): [#31109](https://github.com/sgl-project/sglang/pull/31109)

**CUTLASS FP8 blockwise deleted for SM90 / SM100**, SM120 moved to JIT: [#30438](https://github.com/sgl-project/sglang/pull/30438)

**`--fp4-gemm-backend cutlass` is removed** along with the in-tree NVFP4 JIT kernels, so NVFP4 GEMM now requires FlashInfer. Use `auto`, which picks `flashinfer_cutedsl` on SM100 and `flashinfer_cutlass` on SM120: [#30448](https://github.com/sgl-project/sglang/pull/30448)

**`UnifiedRadixTree` is now the default** for SWA, Mamba and DSA models. A behavior change on those architectures: [#30468](https://github.com/sgl-project/sglang/pull/30468)

from sglang.srt.entrypoints.openai.protocol import Tool ```

from sglang.srt.openai_api.protocol import Tool

Update Prometheus dashboards for new metrics

Review new error response format ## ⚡ Built for speed. Engineered for scale. Production-proven.

Update worker API integrations for UUID-based management