fix(tactic/induction): fix generalisation with complex major premise

[JLimperg]

The induction' and cases' tactics would previously not generalise the goal correctly when called with a complex term (i.e. not just a local constant) as major premise. E.g. the call cases' (n + 1) would fail to replace n + 1 in the hypothesis n + 1 = m, leading to very weird goals.

---