[FEA] port CSV tests from Spark and univocity

[revans2]

Is your feature request related to a problem? Please describe.
Spark does not have a lot of really good CSV tests. They use a third party CSV parser https://github.com/uniVocity/univocity-parsers/ and end up relying on it for a lot of the testing. We should try to port over as many tests as possible from these to our tests to help augment corner case testing. We should concentrate on tests that are for the default settings in Spark, but we should also add in as many tests as possible and verify that we fall back if they do not work, or mark them as xfail with the issues that would fix that problem.