ml4gw introduction

[alecgunny]

• Build monitoring service that archives frames along with clean data in 4096 chunks, deletes stale frames, and increments production version once tests pass
• Tag frames with version of the model used to clean them
• Switch DEEPCLEAN_PRODUCTION channel to permanently be the deepclean'd channel, use _OUT_DQ suffix for data product
• Create central Dockerfile for all microservice projects and add build step to CI workflow