[Enhancement] Add Self Refinement Feature

[psriramsnc]

🚀 SyGra now supports Self Refinement

Summary 🧾✨

This PR adds Self Refinement as a first-class SyGra feature 🔁🤖. This uses newly implemented sygra.recipes.self_refinement recipe as a subgraph inside a task 🧩. Generated datasets that include judge scores ✅📊 and a reflection trajectory 🪞🧠 for each prompt.

Feature implemented: 🚀

• Added self_refinement under recipes.
• Added examples.self_refinement under tasks which uses the above recipe as a sub graph
• Added a new feature doc: docs/features/self_refinement.md describing the Self Refinement workflow 🔄, outputs

Performance impact (if any): ⚡

N/A

How to Test the feature 🧪

Steps for reviewers to verify functionality:

Run the example task and observe generated artifacts ▶️:
Run: python main.py --task examples.self_refinement --num_records 2 🏃‍♂️💻

Observe output JSON under tasks/examples/self_refinement/ (e.g., output_.json) 📦🕒
self_refine.generate_candidate and self_refine.judge_candidate are tracked 📈🧾

Screenshots (if applicable) 📸

N/A

Checklist ✅

• Lint fixes and unit testing done
• End to end task testing
• Documentation updated

Notes 🗒️

N/A