feat: scientific benchmark upgrade + theoretical foundations

fix: correct indentation and complete encode() method in benchmark

fix: align benchmark encoder with app.py 20D approach

docs: bilingual README (EN + ZH) with v2.7 features

feat: v2.7 - corrections P0/P1, benchmark DistilBERT, 73 tests

Initial commit: Tian-Dao-LLM structure with Apache 2.0 license