docs: add tiny training smoke tutorial

[lsr-05]

Summary

• add a step-by-step tiny training smoke tutorial using the existing math dataset loader and reward function
• document GPU/CPU expectations, success logs, checkpoint output, and TensorBoard metrics
• link the tutorial from README, quickstart, CLI reference, and the docs landing page

Closes #41

Test

• \python -m sphinx -b html -D html_theme=alabaster docs docs/_build/html\ (default shibuya theme package is not installed in this environment)

[lsr-05]

#41

[lsr-05]

#41