Avoid Vertex AI Feature Store system Dag import timeout

[viiccwen]

Prevent the Google Vertex AI Feature Store system Dag from loading aiplatform_v1beta1 while Airflow discovers Dags. The package's third-party import chain can exceed the default 30-second Dag import timeout before any task runs.

Defer the Feature Store clients and data-format type until their execution paths need them. Use GAPIC-supported mappings in the system Dag so parsing does not construct protobuf messages, while keeping the existing hook and operator type signatures unchanged.

Update the client mock targets for the deferred imports and cover both client getter paths.

Before/after cold-import comparison in the same Breeze environment:

	Before	After
System Dag import time	30+ seconds (timeout)	10.278 seconds
google.cloud.aiplatform_v1beta1 loaded during parsing	Yes	No

The before result exceeds Airflow's default dagbag_import_timeout of 30 seconds. The after result keeps the Vertex AI Feature Store package out of the Dag parsing process entirely.

• closes: Google Vertex AI Feature Store system Dag exceeds import timeout #69822

---

Was generative AI tooling used to co-author this PR?

• Yes - Codex 5.6 Sol

---

• Read the Pull Request Guidelines for more information. Note: commit author/co-author name and email in commits become permanently public when merged.
• For fundamental code changes, an Airflow Improvement Proposal (AIP) is needed.
• When adding dependency, check compliance with the ASF 3rd Party License Policy.
• For significant user-facing changes create newsfragment: {pr_number}.significant.rst, in airflow-core/newsfragments. You can add this file in a follow-up commit after the PR is created so you know the PR number.

[viiccwen]

cc @guan404ming, @jason810496. 🙌

[shahar1]

Thanks for your contribution!
Got a small deja vu from a similar issue I encountered in the past: googleapis/python-aiplatform#1994
Few comments to handle and we're good to go :)

[shahar1]

# type: ignore[arg-type] will leak into the usage example in the docs, is it possible to avoid it?

[viiccwen]

What if we specify type dict, then remove the ignore flag?

[shahar1]

Could we prevent regressions by asserting on sys.modules? (to ensure that no one re-adds the problematic import)

[viiccwen]

Sure, I'll reference the testing style like test_accessing_bundle_loads_cadwyn(...).

[viiccwen]

Thanks for your contribution! Got a small deja vu from a similar issue I encountered in the past: googleapis/python-aiplatform#1994 Few comments to handle and we're good to go :)

Thx for extra content, PTAL.