Allow character-by-character diffs

[lukaszmoroz]

jest-diff has implemented two modes of comparing multiline strings:

• diffLinesUnified for comparing inputs line-by-line - used by diff, which in turn is used by this package;
• diffStringsUnified, for comparing character-by-character, which is slower but also shows changed substrings within the lines with inverse foreground and background - used by standard equality matchers like toBe.

I have big snapshots with long lines and would prefer to see where exactly in line is the difference.
I suggest changing the diff call to diffStringsUnified or maybe make it configurable if performance is a concern, and set some sensible defaults.

This is for example what jest-diff recommends:

If the input strings can have arbitrary length, we recommend that the calling code set a limit, beyond which splits the strings, and then calls diffLinesUnified instead. For example, Jest falls back to line-by-line comparison if either string has length greater than 20K characters.