Postgres clients exhausted, preventing albumart switch

[kenellorando]

After any user spends long enough time connected on the UI, the following begins to appear in the logs:

postgres    | 2023-02-24 06:55:57.782 UTC [244] FATAL:  sorry, too many clients already                         
cadence     | 2023/02/24 06:55:57 [ERROR][SEARCHBYTITLEARTIST] Could not query DB.                              
cadence     | pq: sorry, too many clients already    

Once this starts to appear, all future requests by that same client (e.g. the same open browser tab) to /api/nowplaying/albumart will fail. Refreshing the browser will fix the issue until enough songs have passed on the radio that enough calls to /api/nowplaying/albumart have been made for things to break down.

This bug can be reproduced by running the stack with at least two unique songs that have different album artwork, then using /api/dev/skip to fast-forward though songs.