Tonic Textual release notes

Fixed an issue where the textual-ml container kept running after its machine learning service failed to start, instead of exiting so that it could be restarted.

Fixed an issue where a redacted PDF page with many replaced values could show one replacement value as garbled characters.

Fixed an issue where a redacted PDF could keep sensitive values in text that the original PDF clipped from view, such as text that overflowed a table cell.

Fixed an issue where some PDF drawings, such as CAD sheets made with Ghostscript, came out of redaction as blank pages.

Fixed de-identification failures for large text files with many detected entities.