Design an extensible request Retryer

Problem

You are building a small library component called a Retryer that wraps an operation (typically a network call) and retries it on failure.

Requirements

1. Default behavior
   • Provide a “default retryer” implementation that retries an operation when it fails.
   • The default should support common retry features:
     • Maximum attempts (or max retries)
     • Delay/backoff strategy (e.g., fixed, exponential)
     • Optional jitter
     • Stop conditions (e.g., do not retry on certain exception types / status codes)
     • Ability to honor a caller-provided cancellation/deadline (e.g., timeout)
2. Extensibility
   • The design must be extensible so that callers can plug in custom behavior without modifying the retryer’s core logic. Examples:
     • Custom backoff policy
     • Custom “should retry?” policy
     • Hooks/observers for logging/metrics
     • Handling server-provided retry hints (e.g., a Retry-After value)
3. Usability
   • The API should be easy to use for common cases, while still allowing advanced customization.

What to deliver

• Propose the public API (interfaces/classes) for the Retryer.
• Explain how the default retryer works.
• Explain how customizations are added (e.g., strategies, composition, configuration objects).
• Call out key edge cases and pitfalls (e.g., thundering herd, non-idempotent operations, deadlines).