palomar: special-case Japanese text indexing using kuromoji (#640)
This is just for posts right now, not profiles (descriptions, display
name, etc).
I'm somewhat confident in the indexing approach (separate duplicate
fields, gated by text detection). And this seems to work ok for simple
cases.
I'm not very confident about all-kanji text and indexing, and mixes of
Japanese and non-english character sets. For example, Japanese and
Korean (CJK), or Japanese and Thai (non-CJK).
One positive thing is that everything is still being indexed in the
regular text fields, using the existing analysis pipeline. So we can
revert the query changes if needed, or improve some corner cases using
query-time-only techniques.
Closes: https://github.com/bluesky-social/indigo/issues/628