fix: QA sweep — 7 bugs fixed, 17 test corrections

Code fixes:
- Sidebar state bleed: reset file store on HomePage mount
- restore-photo: raise error instead of silently skipping colorize
  when DDColor model missing
- PaddleOCR OOM: cap input images to 2048px before OCR inference
- Torch CPU optimization: use --index-url .../whl/cpu on CPU nodes

Test fixes:
- upscale: add exact:true to scale factor button locators
- smart-crop: add exact:true to "Pad to square" locator
- colorize: use regex for model button names (Best/Balanced/Fast)
- enhance-faces: use .first() for ambiguous percentage display
- passport-photo: fix DPI locator, .or() compound, generate fallback
- people: update maxUsers assertions for unlimited (0) default
- automate: "Save Pipeline" → "Save" matching actual button text
- tools.test: add resize to Sharp mock chain for OCR tests

packages/ai/src/ocr.ts

@@ -26,12 +26,15 @@ export async function extractText(
 ): Promise<OcrResult> {
   const inputPath = join(outputDir, "input_ocr.png");
 
-  // Convert any input format (HEIC, AVIF, WebP, TIFF, etc.) to PNG
-  // so Tesseract and PaddleOCR can read it reliably.
-  const pngBuffer = await sharp(inputBuffer).png().toBuffer();
+  // Convert to PNG and cap at 2048px to prevent PaddleOCR OOM on large images.
+  const MAX_OCR_DIM = 2048;
+  const pngBuffer = await sharp(inputBuffer)
+    .resize({ width: MAX_OCR_DIM, height: MAX_OCR_DIM, fit: "inside", withoutEnlargement: true })
+    .png()
+    .toBuffer();
   await writeFile(inputPath, pngBuffer);
 
-  const meta = await sharp(inputBuffer).metadata();
+  const meta = await sharp(pngBuffer).metadata();
   const megapixels = ((meta.width ?? 0) * (meta.height ?? 0)) / 1_000_000;
   const timeout = Math.max(600_000, megapixels * 30 * 1000);