Panic with 'add with overflow' in fastpath instead of OOM

[qinsoon]

There is an issue with this PR: if we try to do any address arithmetic in the allocation fastpath, we may have overflow in the computation, and we do not call out_of_memory() for that case. See the test below. I don't think we have a good way to deal with this, given that we do not want to introduce any check in the fastpath.

mmtk-core/vmbindings/dummyvm/src/tests/issue867_allocate_unrealistically_large_object.rs

Line 27 in 04c0c48

pub fn allocate_max_size_object_after_succeed() {

Originally posted by @qinsoon in #896 (comment)

[qinsoon]

@k-sareen suggested

One alternative, although I don't know how feasible it is, would be to register a custom panic handler and then call either out_of_memory or panic_in_vm (random function) or something to allow the VM binding to crash by itself. We discussed doing something similar in the OOM PR and #746