Add LLM-classified convergence analysis (Section 5 implementation)

Cross-model blinded classification of 205 responses using GPT-4o/Claude/DeepSeek
as rotating classifiers. Results: 71.2% acknowledges_limit, 14.6% incoherent,
7.8% claims_self_grounding, 4.4% attacks_then_retreats, 2.0% conditional.
Convergence rate: 80.7%, Krippendorffs alpha: 0.330.

Force-added past .gitignore so the live Analysis page shows real LLM data
instead of heuristic regex results.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: Jamie Nucho