Fiber Reconciler fixes

[ahti]

This is what I have done so far to make the example (with a little modification) from #523 work.

Currently, things work until returning to the initial state, after which pressing the button will leave an empty screen.

Some comments in no particular order:

• The reproducer target is of course only for debugging and to be removed before merge.
• I've left the logs as normal prints for now, since swift-log doesn't build on wasm rn, we can remove those as well if things get ready before swift-log is fixed.
• Pay no attention to the individual commits, I'll clean those up once things look more mergeable.
• The implementation of ReconcilePass.reconcile(_: , in:, caches:) obviously still needs to be refactored.
• The reconciliation pass currently emits mutations that seem to assume replacing an element transfers any children to the replacement. I've adjusted the DOM and test renderers, but maybe the emitted mutations should be changed instead.
• I tried turning the reproducer into a test case, but that fails one step earlier, while returning to the initial state, due to an out of range index in an insert mutation.
• Another issue I observed: There are cases where a remove mutation is emitted with the wrong parent. The DOMRenderer doesn't use the parent, so it's not an issue there, but it does materialize with the test renderer.
• The parent in the remove mutation is optional, but all places where such a value is created pass a non-nil parent. Is there a reason for this, or can this be turned non-optional?

[carson-katri]

I don't recall specifically, but the parent may be be optional because the root of the hierarchy has no parent and could hypothetically be removed.

[carson-katri]

Tested on a small project and didn't notice any regressions.

[ahti]

I finally found some time and started refactoring ReconcilePass.reconcile (not quite done with that), and also fixed the empty screen when going through all the states of the reproducer and then from .a to .b again (nothing was inserted when a fiber that should have an element had an alternate, but the alternate didn't have an element).

There is still an issue with label content not being updated tho, which I'm going to look into next.

[ahti]

Fixed the stale label content as well.

The only issue left that I am aware of is the .remove mutation with wrong parent. I've extended the test to reproduce that one, too. Oh and the reconcile method could still be dryed a bit further.

I also changed the test renderer so that multiple changed fibers get batched like they do in real renderers. Without that, the test was kind of flaky and would randomly succeed or fail (before adding the part to reproduce the remove bug, that is).

[ahti]

superseeded by #542