Add retry logic for transient connection failures

[thomhurst]

When connecting to the Docker daemon, transient network errors can cause failures that would succeed on retry.

Would be useful to have built-in retry support with:

• Configurable max retries
• Exponential backoff with jitter
• Only retry on transient errors (connection refused, timeout, etc.)

This would make the client more resilient, especially in containerized environments where the Docker socket might not be immediately available.