Add batch scan processing endpoints

[jakebromberg]

Context

PRs #52 and #53 added library code lookup, reviews API, and the Gemini scanner service (single-scan POST /library/scan). This adds batch processing so a DJ can upload multiple vinyl record images in one request, get a job ID back immediately, and poll for results as each item processes asynchronously via Gemini.

The batch endpoint is the primary workflow: DJs scan stacks of records in quick succession (place-tap-flip-tap-repeat), then review results later in bulk.

Schema

• scan_jobs table: UUID primary key, user ownership, status enum (pending/processing/completed/failed), item counters
• scan_results table: per-item results with JSONB context/extraction, status tracking, optional catalog match

Endpoints

• POST /library/scan/batch — accepts multipart images + JSON manifest, returns 202 with job ID
• GET /library/scan/batch/:jobId — returns job status with individual results (ownership-checked)

Client Protocol

Flat image upload + JSON manifest:

• images: all files as a single multer array field
• manifest: JSON string describing item groupings (imageCount, photoTypes, context per item)
• Images consumed in order: first N for item 0, next M for item 1, etc.

Validation

• Max 10 items per batch
• Max 5 images per item
• Max 50 total images
• Image count must match manifest sum