Add Kaggle silver training data, retrain model, reorganize data directory

Extract city_country_map to separate file with 317 cities

Add DATE config and widen entity punctuation cleanup

Make post-processing fully config-driven from resume_config.json

Move post-processing rules from Python to resume_config.json

Remove hand-picked short skills allowlist for generic post-processing

Improve post-processing and add 98 real-world resume training data

Improve structured benchmark analysis and robustness

Clean up training pipeline and add export benchmarks