Support Different Kinds of Substrings

There are different kinds of substrings that can be supported by this library. Currently, the implementation supports substrings with respect to `char`s, but some users will likely want substrings with respect to graphemes instead. Word and sentence substrings could also be supported using the relevant unicode standards.

Altogether, I see the following substring variants being possible:

- CharSubstring: respect to chars, which are unicode scalar values.
- GraphemeSubstring: respect to graphemes (using unicode-segmentation).
- WordSubstring: respect to words (using unicode-segmentation).
- SentenceSubstring: respect to sentences (using unicode-segmentation).
- ByteSubstring: respecting individual bytes (equivalent to slicing the string).

Since we are already looking at a breaking change with #9, the `Substring` trait can be renamed to `CharSubstring` (so there is no ambiguity between substring variants). The unicode-segmentation variants (grapheme, word, and sentence) can be guarded behind a `unicode` feature (or perhaps separate features for each?). The byte variant can be held off on for now, since it really isn't needed and presents issues with properly-formed strings.

This solution will give maximum clarity as to what this crate offers, and will give flexibility for users to choose from the various types of substrings offered.