fix: align app_name with Cloud ML Job ID in deployed environments

[EaCognitive]

Problem

When an AdkApp is deployed to Vertex AI Agent Engine, there is often a mismatch between the local app.name and the Resource ID assigned by the backend. The SDK currently uses the local name for artifact lookups, leading to 404 errors because the deployment infrastructure expects the Resource ID.

Solution

Updated AdkApp._app_name() to check for the CLOUD_ML_JOB_ID environment variable. If present (indicating a deployed environment), the method returns this ID, ensuring consistency with the backend's expectations for artifact storage and lookup.

Changes

• vertexai/agent_engines/templates/adk.py: Added environment variable check to _app_name().

Verification

• Verified that the logic correctly prioritizes the environment variable over the local name when available.

[EaCognitive]

Updated with current main and adjusted the app-name regression test for the upstream ADK version check. Local validation: uv run ... pytest tests/unit/vertex_adk/test_agent_engine_templates_adk.py -k 'test_default_app_name_uses_agent_engine_id or test_initialization_keeps_explicit_app_name or test_default_app_name_uses_fallback_without_agent_engine_id' -> 3 passed. Requesting maintainer review when available.