Retry Logic with Exponential Backoff

[mftee]

Add retry logic with exponential backoff to all Stellar Horizon API calls inside StellarClient. Network failures and transient Horizon errors should be retried automatically rather than immediately failing.

Requirements:

• Create a RetryConfig struct:

  pub struct RetryConfig {
      pub max_attempts: u32,
      pub initial_delay_ms: u64,
      pub max_delay_ms: u64,
      pub backoff_multiplier: f64,
  }

• Implement a generic retry_async<F, T>(config: &RetryConfig, operation: F) -> Result<T> utility function where F is an async closure
• Apply jitter to the delay: actual_delay = delay * (0.8 + random * 0.4) using rand::random::<f64>()
• Retry only on:
  • Network errors (connection refused, timeout)
  • Horizon 5xx HTTP responses
  • Do NOT retry on 4xx responses
• Log each retry attempt at warn! level with attempt number and delay
• Apply this to verify_hash, submit_hash, and get_hash_history
• Make max_attempts configurable via STELLAR_MAX_RETRIES env var (default: 3)

Acceptance Criteria:

• retry_async correctly stops after max_attempts
• Backoff delay increases between retries with jitter
• 4xx responses are NOT retried
• Unit tests for the retry logic using mock operations that fail N times then succeed