Blocks until the configured request (and optional token) budget permits another call, then records the consumption.
Token cost of the upcoming call (0 to skip the token budget).
Signals a downstream throttling error; grows the backoff window (1s → 2s → … → 60s cap).
Clears any active backoff.
Limits the rate of an operation to a configured requests/tokens-per-minute, blocking callers in Acquire until capacity is available. Includes exponential backoff that callers can drive via ReportRateLimitError when the downstream service signals throttling.