try all call kinds in generic type node handling

[metagn]

fixes #23406

Apparently it was intentional that this only checked nkCall but maybe things have changed since, we can see what CI says.

Also n.kind in nkCallKinds is implied already, probably just c.inGenericContext > 0 is enough.

[metagn]

The problem is only encountered in CI with -> so whitelisting nkCommand will probably not break anything. Someone could depend on the difference outside of CI though and checking the node type is still weird. Basically semGenericStmt doesn't capture generic parameter symbols? I think this is the minimal case for tclosure:

template foo(x: untyped): untyped = x

type Foo[A] = foo(A) # `foo A` or just `A` works

proc bar[A](x: Foo[A]) = discard

bar(123) # undeclared identifier: A

Doesn't matter if foo is entirely untyped or not.

Normally in seminst each generic parameter is added manually to scope so the generic parameters not being captured in semGenericStmt doesn't matter. In semtypinst however when typing tyFromExpr nodes the generic parameters are not in scope. I don't know if semtypinst is even aware of generic parameter symbols outside of the binding map, I don't know how generic parameter symbols even enter the scope for regular semtypes behavior.

[metagn]

We should probably remove this if branch altogether and use only the else branch, then fix #23406 for semTypeExpr.