I'd rather have it directly in the source than as a bundle, but I guess it's up to you really.
We do not want to keep anyone from working on it, of course. Just the idea of releasing this also "standalone" as a block allocator package seemed interesting, and still is. The current GJAlloc even includes some C++ code so that GJAlloc can be used instead of std::allocator and so on, where applicable. All of this (or at least the parts concerning Pike) certainly could be managed in the Pike tree and then from time to time made into a .tar.gz release of the allocator. However, what if we moved the allocator development git repository right next to the Pike repository, same ACLs as pike.git? We would also give out access to the github repository readily, of course, but that does not seem like a particularly smart process.
In any case, suggestions on how to achieve a good process for both Pike and standalone releases are more than welcome.