chore: Lazy load Vertex AI dependencies in ADK modules

This is about 35% decrease. This change refactors several ADK modules to import `vertexai` and its submodules only when they are first used, rather than at the top of the file. This improves module load times by avoiding unnecessary imports of large dependencies. Imports are also placed within `if TYPE_CHECKING:` blocks where appropriate.

Co-authored-by: Liang Wu <wuliang@google.com>
PiperOrigin-RevId: 829017293

src/google/adk/evaluation/response_evaluator.py

@@ -18,7 +18,6 @@ from typing import Optional
 
 from typing_extensions import override
 
-from ..dependencies.vertexai import vertexai
 from .eval_case import Invocation
 from .eval_metrics import EvalMetric
 from .eval_metrics import Interval
@@ -30,8 +29,6 @@ from .evaluator import Evaluator
 from .final_response_match_v1 import RougeEvaluator
 from .vertex_ai_eval_facade import _VertexAiEvalFacade
 
-vertexai_types = vertexai.types
-
 
 class ResponseEvaluator(Evaluator):
   """Evaluates Agent's responses.
@@ -68,7 +65,9 @@ class ResponseEvaluator(Evaluator):
       metric_name = eval_metric.metric_name
 
     if PrebuiltMetrics.RESPONSE_EVALUATION_SCORE.value == metric_name:
-      self._metric_name = vertexai_types.PrebuiltMetric.COHERENCE
+      from ..dependencies.vertexai import vertexai
+
+      self._metric_name = vertexai.types.PrebuiltMetric.COHERENCE
     elif PrebuiltMetrics.RESPONSE_MATCH_SCORE.value == metric_name:
       self._metric_name = metric_name
     else: