Ensure correct length calculation for strings in fulltext index

[stebloev]

Let's consider emoji 4️⃣: it consists of 3 seperate code points. How should we treat it when building ngram index for example?

Example from @snaury: country flags like 🇦🇨 (two code points)
One more interesting instance 👨‍👩‍👧‍👦 (7 unicode code points, 25 bytes)

[stebloev]

I checked OpenSearch. It treats unicode symbols as single character when building ngram index.
I'll add tests to test this, if there's no such.

Also if the field type is String we should consider byte as single character instead of unicode code point. I'll write tests for this as well.

[stebloev]

On inner meeting decided not to implement support for byte-strings (all strings are treated as Utf8 in fulltext search)