Revisit token bucket approach with per-IP rate limiting

[whoabuddy]

The example in the docs here seems like a way better approach to rate limiting and we could add a layer of logging to track stats.

[whoabuddy]

Backing up - it's not exactly the same. The problem we want to solve for is having a single location that hits the Hiro API without triggering 429s.

Then we can funnel all requests from the frontend through the cache, and if there is a large amount of fresh data necessary the requests will be spaced out and served accordingly.

So the answer here might be scale with one DO per Hiro API key instead of per-IP rate limiting.

Back to the buckets. Buck-chuckets.