v3 Hamt update

[austinabell]

Summary of changes
Changes introduced in this pull request:

• Updates pointer serialization format, all caching updates had been done previously (!go-interop flag)

Tested against https://github.com/austinabell/go-hamt-ipld/tree/rust/testing

Will wait on #981 and add another method to the API before opening PR for review

Reference issue to close (if applicable)

Closes #959

Other information and links

[cryptoquick]

@austinabell Isn't this where you were going to address that additional case you mentioned on our review call? Also, can you describe why that came to mind, so I can understand better?

[austinabell]

So the point I was thinking through is not relevant to consensus, because the go implementation does not return the previous value. However, thinking through it a third time, I think that in this case a few lines above should be modified a bit.

The case is on set (overwrite = true) and the values are equal, should the value being passed in be returned back. Although it seems weird internally to do this, it would match a HashMap functionality by returning the value that existed. The difference would be that we are not swapping the values, and assuming PartialEq is implemented correctly. I am going to switch it to returning the value that existed, to match a HashMap implementation, but open to just returning the value passed in. This operation should be super quick in every case so I think it's the best decision

[austinabell]

Also, just to be clear, the only case which this would have unintended behaviour is when PartialEq doesn't match serialized bytes equality AND this was the only value changed, the parent roots won't be updated. Definitely nothing to worry about, but worth keeping in mind that PartialEq must match bytes equality to interop in every case with go.