(benign?) memory leak in unused branch of `HM_HH_freeAllDependants`?

[shwestrick]

I believe there is a space leak in the else branch here:

mpl/runtime/gc/hierarchical-heap.c

Lines 1337 to 1341 in a71659b

	if (retireInsteadOfFree) {
	HH_EBR_retire(s, child);
	} else {
	freeFixedSize(myUFAllocator, child);
	}

The space leak is never materializing because at the moment this branch is unused. HM_HH_freeAllDependants is called in exactly two places, both with retireInsteadOfFree=TRUE.

(We're essentially seeing an artifact of a previous memory management strategy.)

Some background:
Every union-find node has an HH struct as a payload, and the HH struct also needs to be freed. For example, when EBR eventually frees a union-find node, the corresponding payload is also freed:

mpl/runtime/gc/hierarchical-heap-ebr.c

Lines 14 to 19 in a71659b

	void freeUnionFind (GC_state s, void *ptr) {
	HM_UnionFindNode hufp = (HM_UnionFindNode)ptr;
	assert(hufp->payload != NULL);
	freeFixedSize(getHHAllocator(s), hufp->payload);
	freeFixedSize(getUFAllocator(s), hufp);
	}

Presumably, the fix should be as simple as adding the line freeFixedSize(getHHAllocator(s), child->payload) to the else branch highlighted at the top of the issue... but it's not easy to test, because of the unused branch. Perhaps a better fix would be to put a mandatory crash with DIE(...) in the else branch to mark it for future inspection.