Add GRPO natural language-to-SQL example with execution-based reward

[NP2241]

Resolves #979

It's a good idea to open an issue first for discussion.

This PR adds a new GRPO example demonstrating natural language to SQL training with an external execution environment. The example shows how to integrate Tunix GRPO with a structured task where rewards are computed by executing generated SQL against a small SQLite database and checking result correctness.

The example is intentionally minimal and mirrors the structure of the existing GSM8K GRPO example to serve as a clear reference for users.

Reference

• GitHub issue: Natural Language to SQL GRPO Example #979
• Existing GRPO example: examples/rl/grpo/gsm8k/

Colab Notebook
N/A — this PR adds an example recipe under examples/ and does not introduce a new public API.

Checklist

• I have added all the necessary unit tests for my change.
  N/A — this PR adds an example and does not modify core library logic.
• I have verified that my change does not break existing code and all unit tests pass.
• I have added all appropriate doc-strings/documentation.
• My PR is based on the latest changes of the main branch.
• I have signed the Contributor License Agreement.
• I have followed Contribution Guidelines.