reordering method redefinitions leads to a different displayed method count

[nsajko]

One order leads to generic function with 2 methods:

julia> struct A end; struct B end

julia> function f(::A) end
f (generic function with 1 method)

julia> function f(::A) end
f (generic function with 1 method)

julia> function f(::B) end
f (generic function with 2 methods)

julia> function f(::B) end
f (generic function with 2 methods)

Another order leads to generic function with 4 methods:

julia> struct A end; struct B end

julia> function f(::A) end
f (generic function with 1 method)

julia> function f(::B) end
f (generic function with 2 methods)

julia> function f(::A) end
f (generic function with 3 methods)

julia> function f(::B) end
f (generic function with 4 methods)

xref #53814

[vtjnash]

Ahahah. I thought #57837 would solve it because it would handle any methods that aren't ambiguous. Except that these methods are ambiguous of course, so it cannot handle them. We instead will indeed require the full n^2 check of type-equality to correctly filter the list.