Perform Schlinkert pruning both forwards and reverse, picking whichever saves more words

[sts10]

Performs Schlinkert prune on given word list in both the direction it was provided and in reverse. Why in reverse? Because (a) that is still a valid way to make a list uniquely decodable and (b) in some cases it saves more words from the original list (BIPS39 English word list is such an example).

BIPS 39 word list as an example

If you perform a Schlinkert prune on the BIPS39 list in its given order, 1911 words are saved. If you reverse all words on the BIPS39 list and run a Schlinkert prune, 2011 words are saved.

With this PR, when a user tells Tidy to do a Schlinkert prune (-K), it now does it in both directions and then actually executes whichever the one saves more words. Cool!

Carefully reversing words with accented characters

Use graphemes!

/// Reverse all words on given list. For example,
/// `["hotdog", "hamburger", "alligator"]` becomes
/// `["godtoh", "regrubmah", "rotagilla"]`
/// Uses graphemes to ensure it handles accented characters correctly.
pub fn reverse_all_words(list: &[String]) -> Vec<String> {
    let mut reversed_list = vec![];
    for word in list {
        reversed_list.push(word.graphemes(true).rev().collect::<String>());
    }
    reversed_list
}

Downsides

It does roughly double the time it takes to run a Schlinkert prune. But I think it's worth it if it does save more words. Tidy isn't built for speed, its built for thoroughness.

[sts10]

	length	prefix-free	suffix-free	Schlinkert-pruned	Reverse Schlinkert-pruned
BIPS39	2048	1999	1881	1914	2011

[sts10]

This felt good enough to merge. Will keep an eye out for bugs!