Implement Retryer

Implement a `Retryer` that wraps a callable and retries on failure under a configurable policy (max attempts, backoff strategy, retryable exception filter). Discussion centers on OOD: how to make the retry strategy pluggable. Requirements Provide a default Retryer(max_attempts, backoff) that: Invokes the wrapped callable. On exception, sleeps according to backoff, increments the attempt counter, and retries until max_attempts is hit. Re-raises the last exception after exhausting retries. Make the design extensible to: Different backoff strategies (fixed, linear, exponential, exponential with jitter). A filter that decides whether a given exception is retryable. A circuit-breaker that gives up after consecutive failures across many calls. A pre- / post-attempt hook (for logging or metrics). Notes Pull BackoffStrategy, RetryableFilter, and Hook out into protocols / abstract classes; the Retryer composes them. Exponential backoff with jitter: delay = min(cap, base * 2**attempt) * (0.5 + random()) — the jitter prevents thundering-herd retry storms. Idempotency caveat: retrying a non-idempotent operation (e.g. POST /charge) requires an idempotency key the caller must provide; surface this in the API or refuse to wrap. Async variant: support both sync and async callables — separate Retryer and AsyncRetryer, or branch on inspect.iscoroutinefunction(fn). The interviewer grades on the extensibility discussion as much as the code: walk through the "open / closed" principle and how a new backoff strategy plugs in without touching Retryer. Preparation Implement the sync Retryer with pluggable backoff and a retryable-exception filter in under 25 minutes. Layer on a Hook interface for on_attempt / on_failure / on_success. Pre-write the async variant on paper so you can sketch it quickly if asked. Be ready to discuss two specific edge cases: (1) catching KeyboardInterrupt (do not), (2) retrying inside a with statement that owns a non-idempotent resource.