Config Management: Evaluation

[AkhileshNegi]

Is your feature request related to a problem?

When running evaluations, users must provide the model name, vector store ID, and prompt as separate inputs each time. This is repetitive and error-prone, especially when running multiple evaluations with the same configuration.

Describe the solution you'd like

Extend the config management feature (added in #430) to support evaluation workflows:

• Allow saving evaluation parameters (model name, vector store ID, prompt) in a config
• Enable running evaluations by referencing a config instead of passing individual parameters

This provides consistency with how configs are used elsewhere and reduces friction when running repeated evaluations.