Document ThinkRetrieve test-time retrieval integration

[itsvaibhav01]

Summary

This PR documents how to use ThinkRetrieve alongside ThinkBooster for test-time-scaling experiments.

ThinkRetrieve complements test-time scaling by retrieving relevant worked examples during reasoning and inserting them into the active reasoning trace. The documentation explains the integration boundary, required configuration, evaluation procedure, and citation information.

Added

• docs/integrations/thinkretrieve.md
• Installation and configuration guidance
• Suggested comparisons against standard reasoning and retrieval baselines
• Reproducibility requirements for model, bank, seed, token budget, and latency
• ThinkRetrieve citation information

This PR is documentation-only. A native ThinkBooster strategy implementation would require adapting ThinkRetrieve to ThinkBooster's step-generator and trajectory APIs, which should be handled as a follow-up implementation.

Citation

@article{thinkretrieve2026,
    title={ThinkRetrieve: Retrieval-Augmented Reasoning Traces for Test-Time Scaling},
    author={Singh, Vaibhav and Ghosal, Soumya Suvra and Gharat, Sarvesh and Pal, Soumyabrata and Narayanam, Ramasuri and Manocha, Dinesh},
    journal={arXiv preprint arXiv:2608.10928},
    year={2026},
    doi={10.48550/arXiv.2608.10928}
  }