Configurable `retry_policy`s

[pront]

Motivation

Currently, retry policies are documented for each sink but are not configurable. Some users may require a more aggressive retry policy. The purpose of the new retry_policy option is to provide flexibility, allowing users to define which status codes will trigger retries.

Flexibility

The default behavior will remain the same, but users will now have the ability to override it globally or on a per-sink basis. This approach enables custom configurations without breaking the existing default behavior.

Implementation

The retry_policy option can be a VRL condition, similar to this one.