Change hashes in Induction so that iteration orders are deterministic

[mezpusz]

Tested with portfolio mode and Discount over UFDTLIA benchmarks (only structural induction, integer induction uses the same structures apart from the induction term set). Diff shows no change in the proofs over multiple runs, and there is no major diff compared to master.

[quickbeam123]

Could you please remind me why storing TermList instead of Term*should already liberate you from hashing on pointers? I think TermList hashes on content which boils down to the pointer+tag for wrapped Term*.

Other than that, could please rename TermHash to SharedTermHash?

[mezpusz]

My bad, I remembered that TermList is properly hashed (wrong) but did not check where content was coming from. I'll change it back and add the same hash function.

[quickbeam123]

Cool!