Pagefind needs to load index chunks based on the stemmed word

[bglw]

Example:

1. Pagefind encounters the text rebase on page A, and rebaseMerge on page B.
2. Page A's word stems to rebas, page B's word stems to rebasemerg.
3. A chunk boundary falls in between these words. So one chunk ends with rebas and the next starts with rebasem.
4. A user is dropped into a search for rebase without any intermediary typing.
5. Pagefind finds the chunks needed for rebase, and loads the rebasem* chunk.
6. Pagefind then stems to word down to rebas and searches.
7. Pagefind does not find page A, as that chunk was not loaded.

Solution: Swap steps 5 and 6 — searching for rebase should load chunks for rebas* instead of rebase*