[Bug] BoxFilter operator swaps width/height kernel arguments

[Jayant-kernel]

Describe the bug
BoxFilter currently applies kernel size as (height, width) instead of (width, height).

When width and height are different, the filter effect is transposed relative to what the UI labels indicate.

To Reproduce
Steps to reproduce the behavior:

1. Upload any image.
2. Add filtering_boxfilter block.
3. Set width = 3, height = 5.
4. Run pipeline.
5. Compare result against expected OpenCV behavior for kernel (3,5).

Code-level repro:

import cv2
import numpy as np
from app.operators.filtering.box_filter import BoxFilter

img = np.arange(8*9, dtype=np.uint8).reshape(8, 9)
out = BoxFilter({"width": 3, "height": 5, "depth": -1, "point_x": -1, "point_y": -1}).compute(img)

expected_wh = cv2.boxFilter(img, -1, (3,5), anchor=(-1,-1), normalize=True, borderType=cv2.BORDER_DEFAULT)
expected_hw = cv2.boxFilter(img, -1, (5,3), anchor=(-1,-1), normalize=True, borderType=cv2.BORDER_DEFAULT)

print(np.array_equal(out, expected_wh))  # False
print(np.array_equal(out, expected_hw))  # True

Expected behavior
width should map to kernel width and height should map to kernel height, so backend should call OpenCV with (width, height).

Environment:
OS: Windows 11 (local repro)
Python version: 3.14
OpenCV version: 4.13.0

Additional context
Likely relevant files:
imagelab-backend/app/operators/filtering/box_filter.py (currently passes (height, width))
imagelab-frontend/src/blocks/definitions/filtering.blocks.ts (labels show width then height)

[Jayant-kernel]

Quick update: I opened PR #126 for this fix.

Could you please assign/review when free? Thanks!