docs: clarify evaluation aggregation

models: add NoStatic legacy two-step comparison

models: add main MIP and ChiUNet checkpoints

models: add NatureIK diffusion UNet checkpoints

models: add main recovery checkpoints

models: add RoboTwin checkpoints

models: add FK weight ablation checkpoints

models: add pure MLP IK checkpoints

models: add NoStatic joint-loss comparison checkpoint