feat: support INT8 ConvRot safetensors

[leejet]

Summary

• Parse ComfyUI int8_tensorwise and ConvRot metadata from safetensors files, including I8 weights and per-row scales.
• Execute INT8 ConvRot linear layers through native ggml operations and reuse quantized activations within a graph.

Related Issue / Discussion

N/A

Additional Information

Performance Comparison

Parameters	ComfyUI	sd.cpp	sd.cpp vs ComfyUI
512×512 CFG=1	6.48 it/s	6.94 it/s	+7.1% faster
1024×1024 CFG=1	2.48 it/s	1.62 it/s	-34.7% slower

Test GPU: NVIDIA RTX 4090
Test model: krea2_turbo_int8_convrot.safetensors

Checklist

• I have read and confirmed this PR follows the contribution guidelines.