configurable words per caption segment, down to 1 #3291

I edit clips in Cap where the caption style needs tight segments and the current segmentation is kind of a dead end for that. Everything comes out at 6-8 words per segment since TARGET_CAPTION_WORDS_PER_SEGMENT and MAX_CAPTION_WORDS_PER_SEGMENT are hardcoded consts (apps/desktop/src-tauri/src/captions.rs:77-78, duplicated in apps/desktop-gpui/src/transcription.rs:1865-1866). One word at a time captions are everywhere in short form right now but that's not the only place I want it, and there's no way to get Cap to do it. You export and redo captions somewhere else.

I'd want a setting for how many words are on screen at once, down to 1. The chunker keeps word timings per caption, so re-chunking an existing transcript should be cheap, no re-transcription needed.

At 1 word per segment the multi-word rules don't apply, so the weak-word boundary extension and the MIN_FINAL_CAPTION_WORDS merge should just be bypassed there.

One thing I'm not sure about: do you want this as a free number (slider or input) or as segmentation presets like the short-form segmentation in #2017?

I can do it as two PRs, first the shared-chunker dedup (#3290), then the setting itself. richie said on discord he's open to it.