Suggestion: test against the properties of error output for portability

[zkamvar]

One thing I'm wondering about for the future is if we should be more specific in these tests to make them more portable (e.g. to other languages). I like snapshot tests as a quick confirmation, but sometimes they can be overwhelming with complex output.

I was thinking that we could test the properties of the error output like confirming that the schemaPath and dataPath, and message match our expectations for the errors. I should probably open an issue with this so it can be reviewed systematically.

Originally posted by @zkamvar in #63 (review)