aws_sagemaker: Stack level tags get added to SageMaker CfnModelPackage causing deployment failure

### Describe the bug

When deploying a stack with tags set at the stack level, cdk tries to apply the tag to sagemaker model package which results in deployment error `Tags are not supported in Model Package versions. Please add them to the Model Package Group`.

The same code deployed without errors before this; the last successful deployment tested was in February 2025.

### Regression Issue

- [ ] Select this option if this issue appears to be a regression.

### Last Known Working CDK Version

2.175.1 (same version no longer works now)

### Expected Behavior

Tags should not be applied to untaggable resources as stated in the `Stack` class docstring:

### Current Behavior

CDK attempts to apply stack-level tags to sagemaker model package.

### Reproduction Steps

```py
from aws_cdk import App, Aws, Stack
from aws_cdk.aws_sagemaker import CfnModelPackage, CfnModelPackageGroup

class CustomStack(Stack):
def __init__(self, scope: App, construct_id: str, tags: dict):
super().__init__(scope, construct_id, tags=tags)

model_package_group = CfnModelPackageGroup(self, "ModelPackageGroup", model_package_group_name="name")

model_package = CfnModelPackage(
self,
"ModelPackage",
model_package_group_name=model_package_group.model_package_group_name,
model_package_version=1,
inference_specification=CfnModelPackage.InferenceSpecificationProperty(
containers=[
CfnModelPackage.ModelPackageContainerDefinitionProperty(
image=f"{Aws.ACCOUNT_ID}.dkr.ecr.{Aws.REGION}.amazonaws.com/test:latest",
)
],
supported_content_types=["application/x-recordio"],
supported_response_mime_types=["application/json"],
),
)
model_package.node.add_dependency(model_package_group)
```

### Possible Solution

_No response_

### Additional Information/Context

_No response_

### CDK CLI Version

2.175.1

### Framework Version

_No response_

### Node.js Version

2.175.1

### OS

Linux

### Language

Python

### Language Version

_No response_

### Other information

_No response_

Start by reproducing the provided Python stack with stack-level tags and inspect how AWS CDK applies tags to the SageMaker CfnModelPackage. Done means synthesis or deployment no longer applies tags to the untaggable model package, while supported resources such as the model package group can still receive stack-level tags.

Written by the indexing model from the issue text.