Fix base_model_arn construction to use private hub when SAGEMAKER_HUB_NAME is set

[mollyheamazon]

Issue

When a model package doesn't have hub_content_arn in its container metadata, model_resolution.py falls back to constructing the ARN from hub_content_name and hub_content_version. The fallback hardcoded SageMakerPublicHub as the hub name, causing eval jobs to fail with DownstreamServiceUnavailable when the base model exists only in a private hub.

Fix

use SAGEMAKER_HUB_NAME env var (already used throughout the SDK for hub resolution) to determine the hub name and account in the fallback ARN construction. Falls back to SageMakerPublicHub with aws account when the env var is not set, preserving existing behavior for public hub models.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[aviruthen]

Failing integ tests pass on CodeBuild: sagemaker-python-sdk-ci-sagemaker-mlops-integ-tests:7cb7850a-fb3e-4ea9-bf7e-3886f0e6ff40

Run failing tests based on PR check sagemaker-python-sdk-ci-sagemaker-mlops-integ-tests:0ef18d13-f532-426c-a787-7909617ae5bf (build 689) where only these two tests below failed:

tests/integ/test_hyperparameter_tuning.py::test_hyperparameter_tuning_e2e PASSED [ 50%]
tests/integ/test_pipeline_train_registry.py::test_pipeline_with_train_and_registry PASSED [100%]