fix: deduplicate overlapping OCR words and use per-word Y positions in overlay

Backend: Add spatial overlap check (>=50% horizontal IoU) to Kombi merge
so words at the same position are deduplicated even when OCR text differs.

Frontend: Add yPct/hPct to WordPosition so each word renders at its actual
vertical position instead of all words collapsing to the cell center Y.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

admin-lehrer/components/ocr-overlay/useSlideWordPositions.ts

@@ -4,6 +4,8 @@ import type { GridCell } from '@/app/(admin)/ai/ocr-overlay/types'
 export interface WordPosition {
   xPct: number
   wPct: number
+  yPct: number
+  hPct: number
   text: string
   fontRatio: number
 }
@@ -66,6 +68,8 @@ export function useSlideWordPositions(
             const wordPos = tokens.map((t, i) => ({
               xPct: cell.bbox_pct.x + i * fallbackW,
               wPct: fallbackW,
+              yPct: cell.bbox_pct.y,
+              hPct: cell.bbox_pct.h,
               text: t,
               fontRatio: 1.0,
             }))
@@ -77,6 +81,8 @@ export function useSlideWordPositions(
           const wordPos: WordPosition[] = boxes.map(box => ({
             xPct: (box.left / imgW) * 100,
             wPct: (box.width / imgW) * 100,
+            yPct: (box.top / imgH) * 100,
+            hPct: (box.height / imgH) * 100,
             text: box.text,
             fontRatio: 1.0,
           }))
@@ -202,6 +208,8 @@ export function useSlideWordPositions(
           wordPos.push({
             xPct: cell.bbox_pct.x + (bestX / cw) * cell.bbox_pct.w,
             wPct: (tokenW / cw) * cell.bbox_pct.w,
+            yPct: cell.bbox_pct.y,
+            hPct: cell.bbox_pct.h,
             text: tokens[ti],
             fontRatio: 1.0,
           })