fix(types): Resolve 'tfx.' prefixed KFP v2/Vertex AI artifact types to native Python classes

[vkarampudi]

Description

This PR resolves a major type resolution bug (Issue #7849) where standard TFX artifacts produced under Kubeflow Pipelines (KFP) v2 or Google Cloud Vertex AI Pipelines fail to resolve to their native Python classes.

The Problem

Under KFP v2 / Vertex AI (which uses IR-based pipeline execution), TFX artifacts are registered in MLMD using their YAML schema title which is prefixed with tfx. (e.g. tfx.Examples, tfx.Model).

When a downstream Python step queries MLMD, TFX's get_artifact_type_class() utility attempts to match the MLMD type name against native Python classes. Because native classes have TYPE_NAME attributes without the prefix (e.g. TYPE_NAME = "Examples"), the exact match check fails:

    if (artifact_type.name == candidate_type.name and ...):

Consequently, TFX prints a warning ("Could not find matching artifact class for type...") and generates an ephemeral generic Artifact subclass. This is highly brittle and breaks any downstream code performing class checks (e.g. isinstance(artifact, standard_artifacts.Examples)) or using subclass-specific properties/methods.

The Solution

We normalize artifact_type.name by stripping the tfx. prefix before checking for a match with native TFX classes:

    artifact_type_name = artifact_type.name
    if artifact_type_name.startswith('tfx.'):
      artifact_type_name = artifact_type_name[4:]

Testing

• Added a new unit test in artifact_utils_test.py (testArtifactTypeRoundTrip) that constructs a mock KFP-prefixed tfx.Examples artifact type and asserts that it is correctly resolved to standard_artifacts.Examples.
• All artifact_utils_test tests pass successfully in the tfx-py310 environment.