Flask background task scheduler doesn't run if werkzeug is not the WSGI server

[jagraff]

In order to prevent the apscheduler from executing twice in debug mode, hasher-matcher-actioner checks whether WERKZEUG_RUN_MAIN=="true" before starting the scheduler, because in debug mode the Flask service will initialize twice - werkzeug will set that environment variable only for the "outer" process, meaning that if werkzeug is the WSGI server this is a reliable way to only run the scheduler once in debug mode. However, the "production" version of HMA uses gunicorn as the WSGI server. Since gunicorn doesn't set that enviornment variable, the production version of HMA will incorrectly not trigger the scheduler.