fix: add inc for double resets

[anfelor]

@bollu and I also ran into the double reset bug a while ago (sorry for not reporting this earlier -- we didn't have a good MWE and put it off at first). However, we fixed it slightly differently: The reference counting can still be sound if there are two reset instructions on the same variable, but then an inc has to be generated before the first reset.

The underlying reason for the bug is thus that Figure 5 of https://pp.ipd.kit.edu/uploads/publikationen/ullrich19counting.pdf has a wrong reset case, where the O^+ function is not called on resets, even though the variable to be reset should be owned (compare to the Ty-Reset rule of https://lean-lang.org/papers/beans_appendix.pdf). In contrast, Koka used to call O^+ in this case before we switched to drop-guided reuse analysis.

This PR fixes the issue by calling O^+ on reset. This might make the fix in #4028 easier to maintain since it prevents segfaults even if there are two reset instructions on the same variable.

[leanprover-community-mathlib4-bot]

Mathlib CI status (docs):

• ❗ Batteries/Mathlib CI will not be attempted unless your PR branches off the nightly-with-mathlib branch. Try git rebase 42215cc0728fbee0705acf87995a8720ec8bcc0d --onto f53b778c0d4a474383eab709d67db5e12357f39d. (2024-05-20 12:42:14)