Fix: Add optional detailed logging to AgentEvaluator

[kubikb]

This PR introduces a log_detailed_results parameter to AgentEvaluator.evaluate() and AgentEvaluator.evaluate_eval_set() in src/google/adk/evaluation/agent_evaluator.py.

Description

This feature enables users to see detailed, per-invocation results during agent evaluations. When log_detailed_results is set to True, the evaluator will log the actual invocation, expected invocation, score, and evaluation status for each step.

Motivation

Without this change, users only see a final, aggregated evaluation result (either successful or failed through an assert). When an evaluation fails, they don't have a breakdown of the results, which makes debugging difficult. This forces them to "fly blind" or resort to lower-level evaluation methods to get more details, which requires extra work.

While the ADK UI offers detailed result observation, having this capability directly within pytest runs is beneficial, especially for CI/CD environments where the UI is not available. This change improves the developer experience by providing a direct and convenient way to debug evaluation tests from the logs.