Proper ESI rate limit implementation

[tkhamez]

The current implementation attempts not to reach the rate limit, rather than retrying failed requests.

It has only limited information about the available tokens and makes some assumptions to calculate waiting times (see Neucore\Service\EsiClient::calculateRateLimitWaitTime()).

@tkhamez
When things fail after the request is accepted, logs usually don’t tell the full story of what actually ran.

Seeing the full execution timeline (latency, retries, failure stage) often makes the root cause obvious. If you can share the JSON payload, I can pass it through AI Badgr and check where it failed.