Small Atomics changes to reduce duplications

[anba]

Here are possible changes to avoid some duplications in the Atomics section.

Changes:

• Partially reverts f1239ca, because Atomics.wait actually only needs the GetValueFromBuffer bits from Atomics.load.
• I've replaced the boolean parameter in ValidateSharedIntegerTypedArray with a List, because some prefer to avoid boolean parameters, and because the List approach allows to write a more concise algorithm.
• The buffer address computation was duplicate in multiple operations, we can avoid this by moving it into the ValidateAtomicAccess operation.

Applies on top of #834, but I wanted to keep possible refactorings separate from bug fixes. And I've directly included the changes from #820 in 62ac9b4.

[littledan]

@syg @lars-t-hansen What do you think of these editorial changes?

[lars-t-hansen]

I think this PR should wait until #1127 is resolved since that will likely rearrange some of the relevant bits. I'm not against it, but I'd probably prefer to re-evaluate changes after that PR lands.

(Personally I doubt that changing from a bool to a list of allowed types clarifies anything, but that's mostly a matter of preference and more for TC39 to decide.)

[ljharb]

@lars-t-hansen @anba now that #1127 is merged, can this be updated?

[lars-t-hansen]

@anba, I'll wait until you notify me about any changes.

[ljharb]

@anba this can be rebased on top of master now that #2015 has landed, if you're still interested in pursuing the third commit.

[bakkot]

The change to ValidateSharedIntegerTypedArray is pretty stale after the addition of BigInts and #1908, and I think the waitable boolean which is now on ValidateIntegerTypedArray is better than a list of allowed types, so I'm going to close this. If there's still a change someone wants to make here, feel free to reopen (or, ideally, open a new PR).