Test for all characters of sample string to be defined in exemplar_chars

When recently updating the `cu_Cyrl` definitions, I had a feeling that the new sample texts contained marks not present in `exemplar_chars`, but I had bad concentration that day and therefore postponed action.

At the very least this leads to tofu in GF UI if these characters don’t make it into the fonts via glyphsets, in the worst case it means the language definitions and therefore fonts are actually not supporting a particular language.

So I wrote a [check](https://github.com/googlefonts/lang/blob/7912fe067c25261f496e3aa28c188cf86ed50431/tests/test_data_languages.py#L284) for it (separate branch).

It’s currently a bit hacky because it removes a bunch of mostly punctuation characters from the sample texts because it yielded too many false positives, but ultimately these are actually not false positives but should actually be present under `exemplar_chars.punctuation`.

So I didn’t make a PR yet because this needs to be thoroughly discussed first.

Current result of the test: `130 failed, 245 passed, 1307 skipped`