Regex::find_from_utf16/ecs2 incorrectly matching against bytes instead of u16 words

[jedel1043]

The find functions for UTF16 and ECS2 strings are incorrectly matching against individual bytes instead of whole u16 words in some cases.

Reproducer

use regress::Regex;

fn main() {
    let input = "赔".encode_utf16().collect::<Vec<_>>(); // U+8D54
    let re = Regex::new(r"[A-Z]").unwrap(); // 0x41 - 0x5A

    let matched = re.find_from_utf16(&input, 0).collect::<Vec<_>>();
    dbg!(matched.is_empty()); // false

    let matched = re.find_from_ucs2(&input, 0).collect::<Vec<_>>();
    dbg!(matched.is_empty()); // false
}

In this case the regex [A-Z] is interpreting "赔" as [0x54, 0x8D], which matches against the "T" character.

[ridiculousfish]

Yikes, great find.