api/v2/lumens/ isn't built to be horizontally scalable

[piyalbasu]

Currently, the endpoint api/v2/lumens/ relies on cached data that is stored in memory. This creates an issue when we have multiple instance in kubernetes running the app. Depending on which image you hit, you'll get a slightly different payload as they (likely) all started running at different times and are running on different schedules.

Thoughts on a solution:

• Do we want to save cached data in a single database and use that as our source of truth? (The current postgresdb in the repo has the above problem of multiple instances)