[codegen] support optional fields with different conditions

[cmyr]

Currently the mechanism we use to support optional fields only supports versions, and assumes that a higher version of a table always contains all the fields of previous versions.

This breaks with post, where 3.0 has fewer fields than 2.0. In addition, there have been a few other instances where we have something like this; for instance in lookup subtables, the markFilteringSet exists conditionally on the appropriate bit being set in the lookup flags.

Overall this would give us more flexibility in codegen, and save us hand-writing code.

That's actually an interesting complication for codegen, this is the first time where we will have a lower version of a table (2.0) that has more fields than a higher version (3.0). We make a bunch of assumptions about fields only being added with versions, so I'll have to think about this some.

Originally posted by @cmyr in #576 (comment)

[dfrg]

This also applies to CFF INDEX objects. If the size is zero, the remaining fields are elided. Like lookups, we’re currently betting that these don’t exist at the end of a table, otherwise we’ll fail to parse.

[cmyr]

support for fields conditional on a flag is added in #937