Add Inference Providers adapter: DeepSeek V3/R1 and Kimi K2 via HF_TOKEN

Speed up CPU demo: model caching, prewarm, live progress, decay budget

Sync Space to verified v10 framework; pin gradio 6.16.0 everywhere

Update framework/sector_sensitivity.py

Update framework/mea_distance.py

Update framework/bias_v2.py

Initial commit