Implement the `CharNormalizer` trait on the `LowercaseNormalizer` struct

[Kerollmops]

It would be much better to directly implement the CharNormalizer trait instead of the Normalizer trait on the LowercaseNormalizer structure. This way, the char map will be created if necessary. Currently, the char map could be missing. You can find an example of how to implement it in the ChineseNormalizer.

charabia/charabia/src/normalizer/lowercase.rs

Lines 15 to 35 in de62ab9

	fn normalize<'o>(&self, mut token: Token<'o>, _options: &NormalizerOption) -> Token<'o> {
	match token.char_map.take() {
	Some(char_map) => {
	let mut new_lemma = String::with_capacity(token.lemma.len());
	let mut new_char_map = Vec::with_capacity(char_map.len());
	let mut s = token.lemma.as_ref();
	for (orig_len, new_len) in char_map {
	let (chunk, tail) = s.split_at(new_len as usize);
	s = tail;
	let lowercased_chunk = chunk.to_lowercase();
	new_char_map.push((orig_len, lowercased_chunk.len() as u8));
	new_lemma.push_str(&lowercased_chunk);
	}
	token.lemma = Cow::Owned(new_lemma);
	token.char_map = Some(new_char_map);
	}
	None => token.lemma = Cow::Owned(token.lemma().to_lowercase()),
	}
	token
	}