Selector optimizations

[bobnik]

Selector Optimizations

This PR stabilizes Redux selector references to prevent unnecessary React re-renders, and extracts common selector patterns into shared utilities. The biggest example of this is we no longer re-render the pattern when a user changes the name of a shape.

When unrelated state changes occur, selectors that aggregate data were returning new array/object references even though the contents were identical. This triggered unnecessary downstream recomputation and React re-renders of the canvas.

We now use two memoization strategies for key selectors:

1. Deep input comparison (createDeepEqualSelector) - Skip computation when inputs are structurally equal
2. Deep result comparison (createResultEqualSelector) - Return cached reference when output is structurally equal

For small arrays of IDs (selectVisibleLayerIds, selectAllImageIds), the deep comparison cost is negligible. For larger vertex arrays, the comparison cost is still much lower than the downstream React/canvas re-render cost it prevents.