feat(koptcontext): Add getAllPanelsFromPage() for panel detection

[vmanojkumar]

Add a new method to detect all panels on a page using Leptonica's connected component analysis. This enables panel-by-panel navigation for comics and manga in KOReader.

Returns an array of panel bounding boxes {x, y, w, h} for panels that are at least 1/8 of page dimensions (filtering out small elements).

---

This change is 

[Frenzie]

Couldn't this be simplified to something like this (untested both functionally and performance-wise, regard as pseudocode)?

Suggested change

	for box in boxaIterBoxes(bb) do
	local pix_tmp = _gc_ptr(leptonica.pixClipRectangle(pixs, box, nil), pixDestroy)
	local w = leptonica.pixGetWidth(pix_tmp)
	local h = leptonica.pixGetHeight(pix_tmp)
	-- check if it's panel or part of the panel, if it's part of the panel skip
	if w >= img_w / 8 and h >= img_h / 8 then
	local box_x, box_y, box_w, box_h = boxGetGeometry(box)
	table.insert(panels, {
	x = box_x,
	y = box_y,
	w = box_w,
	h = box_h,
	})
	end
	end
	for box_x, box_y, box_w, box_h in boxaIterBoxGeometries(bb) do
	if box_w >= img_w / 8 and box_h >= img_h / 8 then
	panels[#panels + 1] = { x = box_x, y = box_y, w = box_w, h = box_h }
	end
	end

[Frenzie]

Let's keep that spread out so it doesn't show as changed in the diff. I just randomly wrote it on one line without thinking about anything like that. :-)