0.6.14

Smart masking (--smart) is about twice as fast and catches more: it now uses up to 4 cores (stepping down to 3 after 10 minutes and 2 after 30 minutes of non-stop work, so a long job does not overheat a fanless Mac) and reads every text with two on-device models — the OpenAI privacy filter and a compact open-source model (Apache-2.0) — whose finds are merged. Pictures and screenshots up to 100 MB are masked on a Mac (was 25 MB). Requires the masking engine omna-pii-mask 0.2.17.