[RFC] Fix potential seqfault in testing for ambiguity

[vchuravy]

fixes #24460

jl_gf_invoke_lookup seqfaults on a UnionAll since that has no field parameters,
I think it was originally intended to just take a tuple type.
Base.rewrap_unionall only returns a UnionAll if m.sig is a UnionAll so I suspect
that this codepath was just never tested with a UnionAll

cc: @timholy

[vchuravy]

@timholy this turns your assertion error into:

detect_unbound_args in Base and Core: Test Failed at /home/vchuravy/src/julia/test/ambiguous.jl:283                                                                                                       
  Expression: need_to_handle_undef_sparam == Set()                                                                                                                                                        
   Evaluated: Set(Method[broadcast_similar(f, ::Base.Broadcast.MatrixStyle, ::Type{Bool}, inds::Tuple{AbstractUnitRange,AbstractUnitRange}, As...) where ElType in Base.Broadcast at broadcast.jl:217, bro
adcast_similar(f, ::Base.Broadcast.VectorStyle, ::Type{Bool}, inds::Tuple{AbstractUnitRange}, As...) where ElType in Base.Broadcast at broadcast.jl:215, broadcast_similar(f, ::Base.Broadcast.DefaultArra
yStyle{N}, ::Type{Bool}, inds::Tuple{Vararg{AbstractUnitRange,N}}, As...) where {N, ElType} in Base.Broadcast at broadcast.jl:202, broadcast_similar(f, ::Base.Broadcast.ArrayConflict, ::Type{Bool}, inds
::Tuple{Vararg{AbstractUnitRange,N}} where N, As...) where ElType in Base.Broadcast at broadcast.jl:207]) == Set{Any}()

[timholy]

I will have a look in the next few days. Many thanks for tackling this!

[JeffBezanson]

I'll see if I can get a better fix working (making invoke fully support UnionAll types).

[JeffBezanson]

replaced by #25495