fix(analysis): hold the OCR permit for the full blocking inference

The OCR dispatcher acquired a borrowed semaphore permit, then ran the
CPU-bound inference in `spawn_blocking`. On cancellation (graceful
shutdown) the future dropped the permit while the detached blocking task
kept running, briefly oversubscribing the blocking pool past
ANALYSIS_WORKERS.

Extract `run_ocr_blocking`, which acquires an *owned* permit and moves it
into the blocking task so the concurrency slot's lifetime matches the
actual work. Covered by a cancellation-invariant test asserting the
permit stays held after the caller future is aborted.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

frontend/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mangalord-frontend",
-  "version": "0.93.5",
+  "version": "0.93.6",
   "private": true,
   "type": "module",
   "scripts": {