Spec tests with memory offsets

[anuraaga]

While working on implementation of atomics, I noticed that the spec tests from here passed despite the logic being trivially broken by ignoring const offsets on memory addresses. While I don't know what the coverage goal of spec tests is, const offsets seem like they may be worth having tests for to make it easier to write correct implementations.

[dschuff]

I think it would definitely be worth adding test coverage for this. Did you end up writing any that you'd be willing to contribute?