Fuse semantic segmentation mask into the trimap

Wire BiRefNet into the pipeline (lazy-loaded) and build the trimap from the
semantic subject mask refined by the green key, used when segmentation is
enabled. generate_trimap (chroma-only) stays as the no-segmentation fallback.

Authority split: the mask decides subject topology (keeps green-tinted hair as
foreground, drops see-through holes), the bright-green key sharpens the screen
boundary, ViTMatte refines a small protective band. Background follows the mask
directly so interior holes survive; the band is small so it does not swallow
finger gaps.

End-to-end on the samples: green hair preserved, finger-gap green 99%
transparent (BiRefNet marks it background; was 100% foreground before), leg
edges smooth -- topology/colour/edge problems local cues could not solve, with
no hard alpha surgery.

- SegmentationSettings wired into PipelineSettings/config; --device propagates.
- TrimapSettings: seg_core_threshold, seg_loose_threshold.
- New debug output seg_mask.png.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

bgfilter/settings.py

@@ -25,6 +25,9 @@ class TrimapSettings:
     fg_safe_radius_ratio: float = 0.006
     min_unknown_radius: int = 4
     min_fg_safe_radius: int = 2
+    # Semantic fusion thresholds (used by fuse_trimap when segmentation is on).
+    seg_core_threshold: float = 0.60
+    seg_loose_threshold: float = 0.15
 
 
 @dataclass(frozen=True)
@@ -94,3 +97,4 @@ class PipelineSettings:
     foreground: ForegroundSettings = ForegroundSettings()
     despill: DespillSettings = DespillSettings()
     model: ModelSettings = ModelSettings()
+    segmentation: SegmentationSettings = SegmentationSettings()