Add BEN2 and BiRefNet-HR as selectable Remove Background models

BEN2 becomes the new default local backend (clean cutouts, strong on
hair/fur edges), with BiRefNet-HR available as a high-res/print
alternate and U2Net kept as the lightweight fallback. Both are
MIT-licensed and download weights from HuggingFace on first use
(cached via the existing hf_cache bind mount), unlike U2Net/SAM which
need an explicit download script.

- config: new BG_REMOVAL_MODEL setting (default "ben2")
- tools.py: remove-background-base64 now tries local backends in
  order (request.model override > BG_REMOVAL_MODEL > ben2/u2net),
  falling back to rembg's birefnet-general session as a last resort
- requirements.gpu.txt / Dockerfile.gpu: add ben2 + transformers deps
  needed for the new backends, with a build-time smoke test for ben2
- frontend: model dropdown in the Remove Background dialog, threaded
  through api.js to the new request field

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ro4PwQKvSc3CH19LSN21Ht

backend/requirements.gpu.txt

@@ -31,3 +31,14 @@ sentencepiece>=0.2.0
 # Install post-container-start if needed:
 #   pip install xformers --index-url https://download.pytorch.org/whl/cu121
 # xformers
+
+# Background removal — BEN2 (default, clean cutouts/hair) + BiRefNet-HR
+# (high-res/print alternate). Both MIT-licensed. Verified against upstream
+# source: neither requires torch>=2.5 despite the BiRefNet repo's own
+# requirements.txt floor — that pin is for its training/eval scripts, not
+# the inference path used here. Weights download from HuggingFace on first
+# use (cached via the hf_cache bind mount, same as the diffusion models).
+ben2 @ git+https://github.com/PramaLLC/BEN2.git
+timm>=1.0.10
+einops>=0.6.0
+kornia>=0.7.0