Sync from GitHub main: parser fixes (SUB-TOTAL, 2-digit qty, barcode strip), ISO dates, conditional second OCR pass, non-blocking model load, OCR_SHARED_SECRET gate + 15MB cap, TIMING behind env var

Reconcile with GitHub repo (diverged since 2026-05-20)

fix(logging): use print(flush=True) for TIMING lines so they appear in HF logs

fix(ocr): run warm-up in background so server accepts connections immediately

chore(ocr): add per-step timing logs to identify bottleneck

fix(ocr): warm up inference engine at startup and add /warmup endpoint

feat: initial OCR server deploy