fix(iOS, Stack): Fix for reattaching Screens when swiping back quickly

[t0maboro]

Description

This PR introduces logic to collect only active (non-dismissed) screen controllers coming from JS state.

When a screen gets dismissed on the native side, its state is sent asynchronously to the JS. In scenarios where multiple screens are dismissed quickly, this can result in stale updates being sent from JS to native. These outdated updates may reference screens that are no longer part of the native view hierarchy.

Since view recycling is disabled, we do not expect the RNSScreenView that was dismissed to be reattached; a new instance should be created when we push an update from JS. Then the new controller instance will be initialized, so these associated with detached screens won't be reused again. Therefore, we can safely filter them out from JS state on the native side. On the JS side, the state will stabilize at some point and align with native, once we receive all asynchronous events.

@implementation RNSScreenView {
...
- (void)initCommonProps
{
  _controller = [[RNSScreen alloc] initWithView:self];

To address this, we now filter out any dismissed controllers from the list received from JS.

Note

I'm leaving the previous behavior as the default, and I'm adding iosPreventReattachmentOfDismissedScreens flag that can enable this fix. We're right before a very important release, and this PR is touching the very basic logic; therefore, we don't want to have this fix enabled by default for now. You can opt-in by setting featureFlags.experiment.iosPreventReattachmentOfDismissedScreens = true. Any feedback would be valuable here.

Closes: #2559

Changes

• Added an attribute to mark controllers that were dismissed.
• Added logic for filtering out screens dismissed natively from JS state.

Before & after - visual documentation

Before	After
before.mov	after.mov

Test plan

Added Test 2559.

Checklist

• Included code example that can be used to test this change.
• Updated / created local changelog entries in relevant test files.
• For visual changes, included screenshots / GIFs / recordings documenting the change.
• For API changes, updated relevant public types.
• Ensured that CI passes

[kkafar]

Looks good overall, just as we talked. I have one question though, leaving it below 👇🏻

[kkafar]

Since this change impacts rather core & crucial functionality of stack, I'd like it to be put behind a feature flag or delay its rollout, especially that we plan to release a next version soon & it might potentially end up in SDK 55. By allowing to control this change through JS we'll allow to patch it much more easily.

I'd suggest adding a "experiment" feature flag, e.g. iosPreventReattachmentOfDismissedScreens and consume it directly on screenstack component.

We can make this flag a noop in two or three minor versions, but next minor should be possibly stable.

What do you think?

[t0maboro]

Since this change impacts rather core & crucial functionality of stack, I'd like it to be put behind a feature flag or delay its rollout, especially that we plan to release a next version soon & it might potentially end up in SDK 55. By allowing to control this change through JS we'll allow to patch it much more easily.

I'd suggest adding a "experiment" feature flag, e.g. iosPreventReattachmentOfDismissedScreens and consume it directly on screenstack component.

We can make this flag a noop in two or three minor versions, but next minor should be possibly stable.

What do you think?

8e0f9b3 + c6e5f1c

[kkafar]

We could consider clearing the state in didMoveToParentViewController in case the controller is reattached. Beside that, looks good.

[t0maboro]

We could consider clearing the state in didMoveToParentViewController in case the controller is reattached. Beside that, looks good.

c9e923b