fix(vertexai): emit cache tokens (#4263)

Author: dvirski

packages/opentelemetry-instrumentation-vertexai/opentelemetry/instrumentation/vertexai/span_utils.py

@@ -311,3 +311,13 @@ def set_model_response_attributes(span, llm_model, token_usage):
             GenAIAttributes.GEN_AI_USAGE_INPUT_TOKENS,
             token_usage.prompt_token_count,
         )
+
+        cached_content_token_count = getattr(
+            token_usage, "cached_content_token_count", None
+        )
+        if cached_content_token_count is not None:
+            _set_span_attribute(
+                span,
+                GenAIAttributes.GEN_AI_USAGE_CACHE_READ_INPUT_TOKENS,
+                cached_content_token_count,
+            )