feat(compressors): add OWLv2 bbox crop to HashPipeline and refactor image utilities

- Add Owlv2ForObjectDetection and Owlv2Processor imports to model_loader
- Refactor load_dino_model to return tuple of processor and model
- Rewrite generate_proposals_batch to group images by bbox count for efficient batching
- Add _normalize_single_bbox_list helper for bbox normalization
- Update verification.py to use new pipeline architecture with detect/segment/filter/crop steps

mini-nav/compressors/model_loader.py

@@ -6,6 +6,7 @@ import torch
 from transformers import (
     AutoImageProcessor,
     AutoModel,
+    BitImageProcessor,
     Dinov2Model,
     Owlv2ForObjectDetection,
     Owlv2Processor,
@@ -35,7 +36,7 @@ def load_sam_model(
 
 def load_dino_model(
     model_name: str = "facebook/dinov2-large",
-) -> tuple[AutoImageProcessor, Dinov2Model]:
+) -> tuple[BitImageProcessor, Dinov2Model]:
     device = get_device()
 
     processor = AutoImageProcessor.from_pretrained(model_name)