refactor(swingset-liveslots): stateShape causes accessor sharing

[erights]

When a kind is provided with a stateShape, we know exactly what the names are for all of the state fields across all instances. For that case, it is wasteful to create accessor properties per instance. So instead, when there is a stateShape, this PR will instead install similar accessors on a shared statePrototype object. Each state object is them a frozen empty object that inherits from that statePrototype. As with a shared prototype for methods, to make this work, we also introduce a separate stateToInnerSelf weakmap, so the accessor can get the correct innerSelf from its this, which is the empty state instance itself.

See also
#7289
#6693
#7138

[warner]

@erights That's really cool, and I look forward to using it. That file, however, is about to be changed drastically by #7138 , and I'm wondering if you could instead look at the target branch (6693-vo-lru-cache) and see if you can make the corresponding change to it instead? I replaced contextMap with a provideContext function, and innerSelf disappears entirely. I think the same approach can be made (and it should be a good bit simpler), but if we land it first, I'm going to spend days trying to understand how it works and changing my PR to match, whereas I bet you could make the same changes to my branch pretty quickly.

I'm going to be rebasing my branch heavily as I get it ready for review and landing (still fighting through some unit test changes now). If you make a branch based on the current 6693-vo-lru-cache, I'd be happy to manage any updates or rebasing that your branch needs as I make changes to my branch.

[erights]

@warner understood. Good plan. I've converted this to draft at least until I rebase and reconcile with #7138

[warner]

@erights 7338-store-state-shape is the branch to base your work on, it's still in review but barring changes it's what trunk should look like in a few days

[FUDCo]

Thatsa very nice.

[erights]

Closing in favor of #7444