- place all the eval { require ... } or plan skip_all in BEGIN {} blocks

 - using `Test::More` module (no hand-rolled TAP or `Test` module)

 - `use strict; use warnings` everywhere

 - remove use of `$a`, `$b` lexicals (reserved for `sort` function)

 - using `Test::Exception` (lives_ok, dies_ok, etc.) rather than the `eval { $code }; ok $@` combination

 - make each individual test run in its own scope
   e.g.
   ```perl
   {
 	my $test = ...
 	ok( ... );
   }
   {
 	my $test = ...
 	ok( ... );
   }
   ```
   This makes tests not dependent on previous code which means that they can be
   moved around without worrying about state. In the future, these should be setup
   as labelled subtests.

 - Using `all approx( ... )` instead of a custom `tapprox( ... )` in each file.

 - update TODO to remove the test refactoring items