feat(web): dictionary-based word breaking 🔬

Predictive text: dictionary-based word-breaking (as in many SE Asia languages)

* The one I’ve given Makara so far is a hack that can definitely break.
* Definitely easier than attempting morphological modeling
* Potential lead: https://www.geeksforgeeks.org/word-break-problem-dp-32/
* It’s not that simple to import, given our sliding context window… but is definitely a great starting point.
* Other possible leads: just search “dictionary-based wordbreaking.”
* Marc has mentioned that ICU has existing implementation for a number of SE-Asian scripts. Especially for C; may not for JS.