Modify Script matching algorithm

First manually match script code based on provided scripts in the
language tag before using UScript to derive script code from tag
and matching that.
Corner case: Uscript.getCode() hard codes script for "ja..." tags
to Japenese, Hirangana and Han scripts even if we provide explicit
"Latn" in the language tag. IME can explicitly say "Latn" in language
tag to allow any KCMs with Latn support and handle composition to
Japense on IME side. So, if explicitly mentioned we should respect
the script code provided.

Test: atest KeyboardLayoutManagerTests
Bug: 272200660
Change-Id: Ic0ffd3aa57d83b7fd8c1fb66ec12581f90c1deaf