Test data for comparison with FreeType

There are several test cases in `read-fonts` and `punchcut` where the condition is “produces results identical to FreeType.” The expected output is currently hand curated which is not ideal. We’d like to be able to generate this programmatically to both ensure consistent data and to facilitate adding new test cases.

I see two potential solutions to this problem:

1. Take on the `freetype` crate as a test only dependency in these projects. Use this inline to generate expected values on the fly.

2. Add a new crate to the repo that is used to generate the test data which is then included in the relevant places.

I’m not sure if 1 violates our new dependency rule. I suppose we should clarify if build and test dependencies are subject to it.

Start by inspecting the comparison test cases in `read-fonts` and `punchcut` and their hand-curated expected outputs. Resolve whether to use the `freetype` crate as a test dependency or add a generator crate, including the dependency-rule question. Done means the relevant test data is generated consistently and supports adding new cases.

Written by the indexing model from the issue text.