Read config from inside inputs, not top-level request body

Support config-driven mode: automatic or point-prompt segmentation

Lower points_per_side to 16 to cut CPU inference latency

Revert mask generator density bump, keep on CPU-friendly settings

Filter automatic masks for trench-like shapes

requirements: sam2 to segment all objects

fix: hardcode model ID, sam2 from_pretrained does not accept local path

requirements: sam2 from source only

add custom endpoint handler for SAM2.1