Organize samples by the matcher

[spenserblack]

This is roughly how linguist does it. They put samples that should be matched by filename in samples/LANGUAGE/filenames/ (extension tests just go in samples/LANGUAGE/).

On a slightly related note, perhaps the "from string" method on Language should be replaced with reversing the name to the language for samples tests, with a pub(crate) method to reverse the variant name to the variant for .gitattributes :thinking: