Various improvements

[vkbo]

This PR:

• Shifts the subtitle index tree view so it starts at 1.
• Adds a clear undo stack step after populating the text editor with the extracted text.
• Removes some regexes that were used to correct wrong capitalisation with the previous OCR settings for Tesseract. These are now more annoying than helpful. In the past, they caught more useful corrections than false positives, but that's no longer the case.