Possible problem in parents upcast generation

[EPashkin]

Some types (GtkCheckButton) have children and only constructors (no trait).
So for it's children ('GtkRadioButton') we better skip parent upcast.
We can just add empty trait or go hard way.

• add boolean flah in analyze::object::Info
• add cache for this in env::Env (or state)
• then when processing parents, check cache for its Info (analyze if not found) and skip upcast.

Currently seen only GtkCheckButton, so empty trait IMHO will do.

[EPashkin]

Forgot to say about GtkScrollbar, and some others. Its have only deprecated children and can be correctly processed with analog of https://github.com/EPashkin/gir/tree/deprecation.
The latest use not too suitable for this module semver and because of this contains hacks.

[gkoz]

Retaining consistency (not skipping over some classes in Upcast impls) seems worth having a few empty traits.

As to the classes with deprecated/useless subclasses, the analyzer might have to recompute the has_children property taking into account the set of widgets present in the toml. If the children aren't present in the config, it's as if there's no children.

[EPashkin]

Agree for counting only "listed" children.