Improve general request performance

[devmount]

What changed?

This is the third (and last for today I promise 😂) PR improving the availability page performance. This one is not scoped to the availability page in particular, but to EVERY single request we made, that needs a database lookup.

We use get_db() as a per-request dependency, which is fine. But we were calling SQLAlchemy's create_engine() via get_engine_and_session() there, causing to open a whole new connection pool without looking for an already existing one.

This PR sets a global engine and session (like we do with the redis instance) and checks, if that already exists before creating a new one.

Why?

Prevent long loading times and confusion by UI sitting seconds in an intermediate state.

How to test

1. Don't check out this branch yet
2. Open the availability page and the web console with the network tab
3. Clear the network tab and hit CTRL+F5
4. See the different http://localhost:5000/* (calendars, external connections, schedule) requests and their timings
   
5. Now check out this branch and repeat steps 2-4
6. See the http://localhost:5000/* requests being faster:
   

Limitations and Notes

This is a very heavy impact change. So please double check if I did that right! Existing tests were running fine though.

Applicable Issues

Related to #1371

Screenshots

No UI changes

[davinotdavid]

LGTM!

QA Log

• Code diff
• Manually tested locally and it does seem to improve performance!