Refactor: flask & APScheduler

[boyska]

The current process model is complex and fragile: it requires being on single-process to maintain job scheduling; depends on paste for no very good reason; uses bottle, which has nothing wrong but is not much used; lose jobs upon restarts.

To address #6 we need better job scheduling, or it just won't do. #25 proposal is good but maybe complex; and it will strictly require a persistent storage for scheduler, which APScheduler doesn't