Hashtable follow-up: Don't expose pointers to internals

[zuiderkwast]

I don't think we should ever be exposing pointers into the structure. Instead, an opaque hashsetPosition can have appropriate accessor methods.

Originally posted by @JimB123 in #1186 (review)

Regarding functions like void **hashtableFindRef() and others that return or work with pointers to internal structures.

I think we need to be careful about this one. Do we really need it? One problem I see is that it doesn't pause rehashing, so the location may be reused by something else or even deallocated before a caller modifies the pointer. There's no protection at all.

I recommend changing this to return a hashsetPosition instead of a void**. Then you can have something like hashsetGetAtPosition(position) and hashsetReplaceAtPosition(position, newElement). You can use epoch information in the hashsetPosition to guarantee safety, and also perform checks like compare(ele1, ele2) to ensure that the structure is still logically correct.

Originally posted by @JimB123 in #1186 (comment)

---

I tried to refactor the code to achieve this, including adding an epoch counter to the hash table structure to be able to detect changes. It turned out to be a non-trivial amount of refactoring. Since it's not a regression and can be done later, I'm moving it to a follow-up issue.