fix: resolve asset staging timing issue and use dynamic region for multi-region StackSets

[spensireli]

Summary

This PR fixes two issues with the StackSetStackSynthesizer that prevent proper multi-region StackSet deployments with assets (e.g., Lambda functions):

1. Asset Staging Timing Issue

The addFileAsset method was using Source.asset() with a local file path that doesn't exist during synthesis. This fix delegates asset staging to the parent stack's synthesizer first, then uses
Source.bucket() to reference the staged asset in S3.

Changes:

• Delegate asset staging to the parent stack's synthesizer via parentStack.synthesizer.addFileAsset(asset)
• Create a lazy reference to the parent's asset bucket using Bucket.fromBucketName()
• Use Source.bucket() to reference the asset from S3 instead of local file path
• This follows the same pattern as ProductStackSynthesizer in aws-cdk-lib

2. Dynamic Region Reference for Multi-Region Deployments

The bucket name was being constructed using this.boundStack.region which hardcodes the parent stack's region at synthesis time. For multi-region StackSets, each target region needs to resolve the
bucket name dynamically.

Changes:

• Changed from Fn.join('-', [this.assetBucketPrefix, this.boundStack.region])
• To Fn.join('-', [this.assetBucketPrefix, Fn.ref('AWS::Region')])
• This allows the bucket name to resolve dynamically at deployment time in each target region

How It Works

When deploying StackSets with Lambda functions or other assets across multiple regions:

1. Assets are staged to the parent stack's asset bucket
2. BucketDeployment copies assets to regional buckets (e.g., my-assets-us-east-1, my-assets-us-west-2)
3. The StackSet template uses ${AWS::Region} intrinsic so each target region fetches from its local bucket

Test Plan

• Existing unit tests pass
• Integration tests pass
• Manual testing with multi-region StackSet deployment with Lambda assets