v1.2.2

Changes since v1.2.1

• fix: dedupe_df() zero-config auto-detection — calling dedupe_df(df) with no config now auto-detects column types and builds matchkeys automatically (previously crashed with __placeholder__ error)
• fix: utf8-lossy encoding for all CSV reads — prevents crashes on files with Latin-1 characters (government data, international names)
• fix: autoconfig classification + blocking safety — dates no longer misclassified as phones, cities no longer misclassified as names, high-null columns skipped for blocking