Move all the various id_sett typedefs into irep.h

[NathanJPhillips]

Call them unordered_id_sett to reflect the fact that these are unordered sets
Add matching id_sett for signed sets of identifiers.

[NathanJPhillips]

Test-gen bump: https://github.com/diffblue/test-gen/pull/1714

[tautschnig]

1. The cleanup seems like a good idea.
2. I don't think id_usett is a great name - can we use unordered_id_sett instead?
3. Any use of id_sett should be audited: it's useful for doing set intersection/set union, but not for much else, because it won't provide the performance of a hash container and doesn't provide lexicographic ordering either (unless irep_idt == std::string).

[smowton]

Agreed re: unordered_id_sett. I wouldn't wade into set -> unordered_set replacements in this PR, since each case requires some searching to check if anything cares about reproducible ordering and/or stable iterators.

[kroening]

This only needs the commit messages reworded, then seems fine.

[kroening]

Considering #2082, I am wondering whether this actually improves readability.

[tautschnig]

I would prefer: merge #2082 and then make good use of the commits in this PR and do sed -i 's/unordered_id_sett/std::unordered_set<irep_idt>/g' (which now actually is possible thanks to all the consistency cleanup done by @NathanJPhillips.