Cache custom-query JSON responses to avoid API rate limits

[diorwave]

No description provided.

[0xSpuddy]

looks good to me. Offered a small waiting behavior fix, but should work well enough as is.

[0xSpuddy]

Should we use sharedCache.group.DoChan instead of sharedCache.group.Do to avoid callers waiting longer than they need to when the leader's http call times out? (the code works as is, this is a small waiting behavior fix)

ch := sharedCache.group.DoChan(key, func() (interface{}, error) {
		// Re-check: a concurrent leader may have just populated the cache.
		if body, ok := sharedCache.get(key); ok {
			return body, nil
		}
		body, err := r.fetchWithRetry(ctx)
		if err != nil {
			return nil, err
		}
		sharedCache.set(key, body)
		return body, nil
	})

	select {
	case res := <-ch:
		if res.Err != nil {
			return nil, res.Err
		}
		return res.Val.([]byte), nil
	case <-ctx.Done():
		return nil, ctx.Err()
	}
}

[0xSpuddy]

Looks cool, and it will be interesting to see if calls go down after the changes. Will merge into main now and then test it on palmito for a day before making a new release with all these pull requests in.