Metis doesn't find an old proof in acceptable time now that higher-order equality reasoning is supported -- tuned proof script to help it

drop support for Vampire's native output format -- it has too many undocumented oddities, e.g. "BDD definition:" lines

use the same code for extensionalization in Metis and Sledgehammer and generalize that code so that it gracefully handles negations (e.g. negated conjecture), formulas of the form (%x. t) = u, etc.

further lower penalty associated with existentials in Sledgehammer's relevance filter, so that "exhaust" facts are picked up

reenabled equality proxy in Metis for higher-order reasoning

added "SMT." qualifier for constant to make it possible to reload "smt_monomorph.ML" from outside the "SMT" theory (for experiments) -- this is also consistent with the other SMT constants mentioned in this source file

reflect option renaming in doc + do not document the type systems poly_preds? and poly_tags?, since they are virtually identical to the non-? versions

unfold set constants in Sledgehammer/ATP as well if Metis does it too

do not pollute relevance filter facts with too many facts about the boring set constants Collect and mem_def, which we might anyway unfold depending on Meson's settings

renamed "max_mono_instances" to "max_new_mono_instances" and changed its semantics accordingly