Major refactoring

[bastoica]

Description

This PR implements shared base classes and reusable primitives for agent validation oracles. It also refactors the artifact-specific agent validation oracles for Acto, Anvil, EqWalker, and Wasabi to use the same unified/standardized code structure. The goal is to reduce duplicate logic, make the oracle code easier to read and maintain, and ensure agent validation runs consistently across artifacts in ArtEvalBench.

Changes

• Implemented 4 base oracle classes corresponding to the 4 canonical stages of the artifact evaluation process (environment setup, benchmark preparation, artifact build, and experiment runs) along shared requirement/check primitives.
• Refactored Acto, Anvil, EqWalker, and Wasabi artifact-specific agent validation oracles to the standardized primitives and base orchestrator flow.
• Created a README that documents the base classes and primitives, explains how to derive new oracle classes, and shows how to use the helper functions; basically this serves as a step-by-step guide for implementing the agent validation oracle part for any new artifact to-be-added to ArtEvalBench.

Testing

Installed the refactored artifacts locally and ran the agent validation oracles as standalone using a main.py runner.

Checklist

• [x ] Tests pass locally
• [x ] Code follows project style guidelines
• [x ] Documentation updated (if needed)