[#17723][fix] Re-path dense Qwen quantization entries

[mihai-chiorean]

Summary

Translate dense Qwen3.5/3.8 MLP entries from the Hugging Face .mlp.{gate,up,down}_proj namespace to the runtime _DenseMlpAdapter .mlp.mlp.* namespace for every quantization algorithm.

Related to #17723.

Motivation

The existing translation was gated on W4A16_NVFP4. Qwen3.8 compressed-tensors checkpoints correctly parse packed W4A4 layers as NVFP4 and later dense layers as rowwise FP8, so neither entry matched the runtime module tree. The modules then silently inherited the unquantized MIXED_PRECISION global config.

Testing

• pre-commit run --files on all changed files
• Focused tests cover NVFP4, W4A16 NVFP4, and rowwise FP8 entries
• Included in a 200-pass integrated Spark run; the one old-image-only failure expects the pre-[TRTLLM-12720][feat] Support nvfp4 w4a16 on sm120 #16511 W4A16 method selection, while current main contains [TRTLLM-12720][feat] Support nvfp4 w4a16 on sm120 #16511
• Full Qwen3.8-27B NVFP4 SM121 smoke selected native NVFP4 GEMM and generated coherent output

Dependencies

The translation itself is independent. Qwen3.8 obtains these per-module entries from #17725.