Converting features from .ufo to .glyphs should allow invalid code

I’m trying to roundtrip a .glyphs file that happens to have outdated classes with some glyphs that are not in the font.

in `def _to_glyphs_features()` (in features.py), > `FeaDocument()` > `self._doc = parser_.parse()` > `featureLib.parser.Parser.parse()` it checks for missing glyphs. But all we want is to segment the .fea file into individual feature blocks. The code should not required to be valid.

Start in features.py at _to_glyphs_features(), then follow FeaDocument() through self._doc = parser_.parse() and featureLib.parser.Parser.parse(). Reproduce the .glyphs roundtrip with outdated classes and missing glyphs; done means the .fea content is segmented into individual feature blocks without requiring valid glyph references.

Written by the indexing model from the issue text.