generics with circular dependency through bound typevar do not resolve signatures correctly

[xhantu]

Describe the Bug

Checked code

Multiple generics with type variables bound to the generics so that circular dependencies are created.
See sandbox link.

Observed behavior

Error `Object of class `object` has no attribute `f` [missing-attribute]

The signatures of the the bound types do not seem to be fully build when the type check is performed.

Which signatures are available seems to be related to the order of the used type definitions and not the order of the definitions of the generics.

Expected behavior

No error.

Further info

Also fails if bound types are not defined explicitly on there own but used directly when defining the type vars. See additional sandbox link.

This may be related to #177

Sandbox Link

https://pyrefly.org/sandbox/?project=N4IgZglgNgpgziAXKOBDAdgEwEYHsAeAdAA4CeS4ATrgLYAEALqcROgOZ0Q3G6UN0AqADroRIpsRh0A4gH0AgnQC8deQG156UgBpVWgLrjmUuQCFldUxq27NpQ%2BgknZAYQsvrOvfbHoAxlCocHCqagAqAGqypogy0bqRrrFyLvqIvnSZdJgwYHSysqwQDAUAFHAwUGC62LGJprp%2BdVEuAJR0ALQAfHQAcrjoMOnoWaN02IRgpa0iY1l%2Bk9MZWTl5UxVV7d19A0Ozc3TEQXC%2BAceW4VHyyQoJLTepw-uZq-mF6MVlG9V0qM23dCadESbU6PX6g2GB1%2BixmIwOCymcOe2VydHWlTAW3BuyhByOwVOgWCdA8iWucXkd2iN1MaWWLzRBSKJVk5Uxuj%2BwKuNX%2BpmxO0hKNGqFhwqyEyRDNRa3ZmzBgr28LmBJO6BA2hAZEouSgpEIDFoUAoAGI6AAFUg6sB6uhoLB4fCAgaQNgAV0oqAYEAGhBEZoAyjApAALBgMYhwRAAemj2t1%2Bt4bGjMHQ0cwuD8cGjfhdEHdnu9A2j6N4vwAbqhoKhsLBnehXR6vT6RrhiEX0HA-egyAwQwMOuWYJQ4C2LEIQABmQgARgATBORCAAL6a1B%2Bb1DgBi0BgFHtOAIJHIy6AA

(Only applicable for extension issues) IDE Information

No response

[grievejia]

Yes I think it's the same issue as #177 -- we resolve typevar bound way too early (when they are defined) which results in cyclic dependencies.

I've spent several days in the past at the problem trying to come up with a working fix but didn't succeed. Perhaps someone with better understanding of the generic instantiation logic might have a better chance at it.