persistent queue

A persisting queue will allow the script to fail and/or be cancelled, and resumed later. It will also make hafas-generate-gtfs work in memory-constrained environments, or with very large areas.

Changes necessary:

1. Specify all tasks (departures fetching, trips fetching) in a JSON-serialisable format. (['fetch-trip', tripId, lineName], etc).
2. Implement an in-memory store for stops, lines & trips (see API below).
3. Switch from queue to a persisting one. (I'd like to try bullmq.) Also mention in the readme that Redis is required.

store API

{
	hasStop: async id => Boolean,
	getStop: async id => Stop,
	putStop: async (id, stop) => {},
	// same with lines & trips
}