Implement a Leaky Bucket Limiter

Design and implement an in-memory leaky-bucket rate limiter. A limiter is configured with: - `capacity`: the maximum amount of queued load the bucket can hold - `leak_rate`: how many units drain from the bucket per second Each incoming request contributes 1 unit to the bucket. When a request arrives: 1. First, reduce the current bucket level based on how much time has elapsed since the last update. 2. If adding the new request would keep the bucket level at or below `capacity`, accept the request and update the bucket level. 3. Otherwise, reject the request. Implement a thread-safe API such as `allow_request()` for use in a multithreaded server. Follow-up discussion: - How would you make the implementation safe under high concurrency? - What synchronization strategy would you use, and what are the trade-offs? - How would you handle time precision, bursty traffic, and clock-related edge cases? - How would you scale this design if many application instances must share the same rate limit?