Support specifying locales instead of hard coding to en_US.UTF-8

This is a follow-up task of IMPALA-2019 where we add the initial support for hive-compatible UTF-8 string functions (turned on by setting UTF8_MODE=true). Currently the locale is hard coded as "en_US.UTF-8" which serves most of the cases. It'd be better if we provide a query option to change it. E.g. for Turkish, the uppercase of "i" is "İ", which can't be served by locale "en_US.UTF-8".