CNG Photo Data Automation Portal Prototype QA

Scope checked:
- SQLite database initialization and migration-safe review columns.
- Filename candidate extraction for demo CNG and odometer readings.
- Manual correction override for entered CNG and odometer values.
- Multipart upload parsing with saved photo payloads.
- Daily CSV export with distance, km/kg, review status, and confidence notes.

Current result:
- python3 test_cng_portal.py passes.
- No external services or paid APIs are required for the prototype.

Production note:
- Replace the filename candidate extractor with Tesseract/OpenCV or an approved cloud OCR provider after the buyer shares real dispenser and odometer photos.
- Keep the manual review status in the production workflow so uncertain reads are corrected before the daily 7 a.m. report is trusted.