Allow Lambda Layers to be updated and exported in a shared stack as long as they are retained

### Name of the resource

AWS::Lambda::LayerVersion

### Resource name

_No response_

### Description

Currently, deploys of Lambda Layer updates fail if they are in a different stack than the function or functions that use them, and the reference to them is exported in the output parameters.

It would be helpful if layers could be defined in a separate stack, their reference exported, and allow for a new version to be created, as long as the old version is retained.

### Other Details

Without this it is impossible to create a shared "library" stack to serve multiple function stacks. Templates become overwhelmingly large when everything is combined into a single template, and using nested stacks masks errors in the referenced stacks.

Alternatively, I would be OK if the error reporting of nested stacks actually reported what failed within the nested stack when it fails

No files, tests, or entry points are identified in the issue, so begin by clarifying the deployment and retention behavior for AWS::Lambda::LayerVersion across separate stacks. Done means a shared library stack can export a layer reference, create a new layer version, and retain the old version without the dependent function stacks failing.

Written by the indexing model from the issue text.