fix: remove hardcoded CPU device in VLM_COCO_Local

[Wanbogang]

Problem

VLM_COCO_Local had self.device hardcoded to "cpu", which means
GPU was never utilized even when CUDA is available. This caused
unnecessarily slow local inference on machines with a dedicated GPU.

# Before
self.device = "cpu"  # GPU ignored entirely

Solution

Replace the hardcoded value with automatic device detection using
PyTorch's built-in torch.cuda.is_available():

# After
self.device = "cuda" if torch.cuda.is_available() else "cpu"
logging.info(f"COCO Object Detector using device: {self.device}")

• If CUDA is available → uses GPU automatically
• If no GPU → falls back to CPU silently
• No breaking changes to existing behavior

Changes

• src/inputs/plugins/vlm_coco_local.py — replace hardcoded "cpu" with auto-detection
• tests/inputs/plugins/test_vlm_coco_local.py — mock torch.cuda.is_available() to ensure consistent CI behavior regardless of hardware

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ All tests successful. No failed tests found.

📢 Thoughts on this report? Let us know!