Add caching to /statistics endpoint

[piprett]

This endpoint is significantly slower than a lot of others, due to doing large database queries. It takes multiple seconds on production, and makes graphing hard since I don't want to downgrade performance. Caching this endpoint in redis would improve the response times, and reduce DDoS risk. An alternative approach would be to keep count in-memory, then increase this number when X happens. This is even more performant, but less reliable.

[piprett]

I had an open PR on the labrinth repo for this, however I have decided to not re-create it in the monorepo. This is primarily because the code was too repetitive, and I think there's likely a better way to do it. I still think it's a valid issue however.

[IMB11]

Closing due to #5897 , @tdgao should consider tanstack caching on the frontend for requests in a close timespan with the same queries.

[piprett]

@IMB11 I believe this issue has not been solved, and will not be solved in the linked PR. The issue is about the global labrinth statistics endpoint, not analytics, not frontend. See the comment you made in the old labrinth PR.

The performance looks to have become somewhat better (infra upgrades?), but still worth server-side caching imo.

$ time curl https://api.modrinth.com/v3/statistics
{"projects":123591,"versions":1104527,"authors":53041,"files":1205368}
real	0m2.035s
user	0m0.042s
sys		0m0.009s

[Prospector]

i don't think this is a significant issue and likely some sort of caching happens in redis? unsure. anyway, if an issue our infra engineer will bring it up