Fix the request pacing for a rate-limited API

An AI synced issues from an API that allows 60 requests a minute with Promise.all(ids.map(fetch)). It fired every request at once, and the integration's token was banned for an hour. The new client sends requests in order on a schedule that respects two limits: how many may start in the same second, and how many may start in any 60 seconds.

An assistant wrote plan_request_schedule(request_count, requests_per_minute, max_burst), which computes that schedule. It is in the editor. It has a bug. Find it and fix it.

Input and output:

• request_count: how many requests to send.
• requests_per_minute: the most requests that may start in any 60 second window.
• max_burst: the most requests that may start in the same second.
• Return the second at which each request starts, in order.

Rules the tests check:

• If any argument is 0 or less, return [].
• Requests start in order, each at a whole second. Request i, counting from 0, starts at the earliest second that meets all of these:
  • it is 0 or later
  • when i is at least max_burst, it is at least 1 second after the start of the request max_burst places earlier
  • when i is at least requests_per_minute, it is at least 60 seconds after the start of the request requests_per_minute places earlier