Prepare Scholar Lens for release

Fix Modal inference startup for Scholar Lens

Fix Modal vLLM CUDA startup

feat: add local inference and submission notes

feat: align app with small-model strategy

Add paper results context and AI paper chat

Finish professional review improvements

Improve search reliability and medium-priority UX

Add input limits and complete clear reset

Secure Modal inference endpoints

Use Mistral Small 3.1 for judge demo inference

feat: add grounded Q&A and faster multi-source search

Fix inference startup and deduplication with Codex-assisted cleanup

[codex] Added Modal AI summarization integration