Fix consuming window insets in system window insets padding

[svastven]

This PR fixes an issue where window insets (like statusBarsPadding()) did not correctly respect or propagate consumed window insets. The core problem was that the InsetsPaddingModifierNode logic was "hidden" from the layout traversal system because it was implemented via delegation.

The Modifier.windowInsetsPadding now applies two separate modifiers which create nodes of two types:

• PlatformInsetsPaddingModifierNode is an InsetsPaddingModifierNode that handles actual layout padding and correct resolve of window insets consumed by ancestors.
• PlatformWindowInsetsPaddingModifierNode is responsible for retrieving platform-specific WindowInsets values. On attachment, it traverses up the modifier chain to find the ancestor PlatformInsetsPaddingModifierNode and establishes a link. It provides the insets back up to this ancestor node.

Fixes CMP-9390 iOS seems to ignore consumeWindowInsets

Testing

Adds skiko tests to WindowInsetsTest

This should be tested by QA

Release Notes

Fixes - Multiple Platforms

• Fix window insets consumption in system window insets padding modifiers.

[MatkovIvan]

Adding both of them works, but kind of suspicious and creates more elements to allocate/iterate/etc 🤔

Just to be informed: what's about these alternatives?

1. Delegate PlatformWindowInsetsProviderNode from InsetsPaddingModifierNode (not visa versa).
2. Using DelegatableNode.visitAncestors(includeDelegates = true)

(1) looks more promising for me. WDYT?