`hooo_dual_rev_neq` should require excluded middle for `A, B`

[Sven Nilsen]

See #620

Since hooo_dual_eq is too strong, a new constructive proof is desirable for hooo_dual_rev_neq.

A such proof exists if excluded middle is required for A, B.
This is required due to or::from_de_morgan being used in the proof.

I think this proof is acceptable, given that the only reference to hooo_dual_rev_neq is in tauto_from_para_transitivity which requires excluded middle.

[Sven Nilsen]

This can be also used to prove tauto_hooo_dual_rev_neq when A, B uses excluded middle.