Investigate the more complex variants of Myers' diff algorithm

[fredemmott]

This might result in better performance, and fewer hunks/more human-friendly diffs may be a side-effect (as it's closer to what other diff implementations use). It will not affect the length of the vec<DiffOp>

For example, diff and git diff produce:

@@ -1,5 +1,9 @@
 function testcase(): void {
-  some_function_with_a_long_name_and_several_args('aaaaaaaa', 'bbbbbbbb', 'cccccccc');
+  some_function_with_a_long_name_and_several_args(
+    'aaaaaaaa',
+    'bbbbbbbb',
+    'cccccccc',
+  );
   some_function_with_a_long_name_and_several_args(
     'xxxxxxxx',
     'yyyyyyyy',

Whereas difflib produces:

 function testcase(): void {
-  some_function_with_a_long_name_and_several_args('aaaaaaaa', 'bbbbbbbb', 'cccccccc');
   some_function_with_a_long_name_and_several_args(
+    'aaaaaaaa',
+    'bbbbbbbb',
+    'cccccccc',
+  );
+  some_function_with_a_long_name_and_several_args(
     'xxxxxxxx',
     'yyyyyyyy',
     'zzzzzzzz',

[fredemmott]

@dakotareier fyi https://blog.jcoglan.com/2017/03/22/myers-diff-in-linear-space-theory/ gives some examples where these variants will give /less/ human-friendly diffs