Non-moving collector is unable to finalize weak pointers

Currently, the non-moving collector is unable to finalize weak pointers under some circumstances due to an interaction between the moving and non-moving collectors' weak pointer marking logic. Specifically, consider a case such as:

  nursery             ┆      non-moving gen.
                      ┆
                      ┆
                      ┆       K
                  ┌───┆────► ┌───────┐
                  │   ┆      │       │
                  │   ┆      │       │
                  │   ┆      │       │
   W              │   ┆      └───────┘
  ┌────────────┐  │   ┆
  │WEAK        │  │   ┆
  │ key        ├──┘   ┆
  │ value      ├──────┆──────► ...
  │ finalizer  │──────┆──────► ...
  └────────────┘      ┆
                      ┆
                      ┆

1. At some point the weak will be promoted into the non-moving generation.
2. At some point later, we will initiate a major collection and W will be moved to oldest_gen->old_weak_ptr_list (by initWeakForGC)
3. We proceed with scavenging and eventually call traverseWeakPtrList, which will eventually call tidyWeakList(oldest_gen)
4. We find that isAlive(w->key) == true as w->key is in the non-moving generation and isAlive necessarily conservatively considers any objects in the non-moving generation as alive
5. We call scavengeLiveWeak on w, which will evacuate &w->key. As w->key is a non-moving object, it is pushed to the update-remembered set (see Note [Aging under the non-moving collector]).

At this point we are in a situation where w cannot be tombstoned by the nonmoving GC as w->key is reachable via the update-remembered set.