turns out we can import with shorter names (and we actually already do that frequently in "From Coq Import")

This avoids ambiguity with P and Q that we were using before for both
uPreds/iProps and indexed uPreds/iProps.

- The direction of big_sepS_later and big_sepM_later is now like later_sep.
- Do not use generated variables in the proofs.

It is doing much more than just dealing with ∈, it solves all kinds
of goals involving set operations (including ≡ and ⊆).

simplify_equality        => simplify_eq
simplify_equality'       => simplify_eq/=
simplify_map_equality    => simplify_map_eq
simplify_map_equality'   => simplify_map_eq/=
simplify_option_equality => simplify_option_eq
simplify_list_equality   => simplify_list_eq
f_equal'                 => f_equal/=

The /= suffixes (meaning: do simpl) are inspired by ssreflect.

The tactic injection H as H is doing exactly that.

Also, specialize the big ops to gmap and gset because that is all that
we are using. For the big ops on sets this also means we can use Leibniz
equality on sets.

Do not use proper explicitly but let setoids handle it.