fix: fix short tokens in getEmojiByShortcode

[nolanlawson]

Fixes #88

The issue here is that shortcodes like smiling_face_with_3_hearts contain a short string, 3, which is not indexed in the search tokens. We also have problems with shortcodes like v (:v:) where there are no usual shortcodes at all.

Of course the easiest solution here would be to create an IDB index on shortcodes. However, this seems like a wasteful use of disk space to me, as 1) getEmojiByShortcode() is not always used, and 2) all shortcodes become tokens that are indexed for searching anyway.

So I'm going to keep the current system of filtering the database using the shortcode broken up into search tokens, while also adding special cases for tokens like 3 or v (which requires a full database scan – reasonable IMO given that so few shortcodes are like this).

[nolanlawson]

In Chrome, even on 6x slowdown, the difference between the full DB scan and the fast lookup is not terrible (~6ms vs ~60ms, 10x slower). Given how rare a shortcode like v is (or a nonexistent shortcode), this seems fine to me.

In Firefox it's equally negligible, <10ms in both cases (no slowdown).

[nolanlawson]

Well actually it can get pretty slow if it has to do a full scan because the shortcode doesn't exist. I may have to look into a batching cursor.

Chrome 6x slowdown on left, Firefox on right