Add optional configurable request timeouts + retry support (429/5xx) to improve client reliability #159
Description
Hi! I noticed berserk currently relies on the underlying HTTP behavior for timeouts/retries. Transient failures like 429/502/503/504 or network hiccups can cause scripts to fail unless every user implements their own retry wrapper.
Proposal:
Add optional support for:
- Configurable default request timeouts (connect/read)
- Retries for transient errors (429/5xx, connection errors)
This could be implemented either via:
- A requests adapter (
urllib3.Retry) attached to the session, or - A lightweight retry wrapper in the request layer.
I’m happy to implement this if the maintainers agree with the approach.
Does this sound acceptable?