…ponses

Implements comprehensive usage tracking for token consumption and execution statistics:

**Token Usage Tracking:**
- Track prompt_tokens, completion_tokens, and total_tokens from LLM responses
- Support both streaming and non-streaming LLM interactions
- Aggregate usage across multiple LLM calls in iterative agent execution

**Execution Statistics:**
- Track iteration count (number of agent execution loops)
- Track message count (total messages in conversation)
- Track tool_calls count (total tool invocations)
- Track failed_tools count (number of failed tool executions)

**Implementation:**
- Created UsageTracker component to aggregate metrics during execution
- Integrated tracker into agent streaming execution (RunWithStream)
- Inject tracker via context in both background and streaming task handlers
- Populate Task.Metadata field with collected metrics on task completion
- Added configuration option AGENT_CLIENT_ENABLE_USAGE_METADATA (default: true)

**Testing:**
- Comprehensive unit tests for UsageTracker with thread-safety verification
- Integration tests for both background and streaming task handlers
- All existing tests pass without modification

**API Changes:**
Task responses now include metadata section with usage statistics:
{
  "task": {
    "id": "task-123",
    "status": { "state": "completed" },
    "metadata": {
      "usage": {
        "prompt_tokens": 156,
        "completion_tokens": 89,
        "total_tokens": 245
      },
      "execution_stats": {
        "iterations": 2,
        "messages": 4,
        "tool_calls": 1,
        "failed_tools": 0
      }
    }
  }
}

Closes #126

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-authored-by: Eden Reich <edenreich@users.noreply.github.com>

…tracking

Add comprehensive example showing how to track and access token usage and
execution metrics through the Task.Metadata field. This example demonstrates:

- Automatic token usage tracking (prompt_tokens, completion_tokens, total_tokens)
- Execution statistics collection (iterations, messages, tool_calls, failed_tools)
- Configuration via A2A_AGENT_CLIENT_ENABLE_USAGE_METADATA environment variable
- Client-side metadata access and display
- Integration with default task handlers

The example includes:
- Server with calculator tool demonstrating tool call tracking
- Client with formatted metadata display
- Complete Docker Compose setup with Inference Gateway
- Comprehensive README with use cases and troubleshooting
- Updated examples/README.md to reference the new example

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Eden Reich <edenreich@users.noreply.github.com>

Co-authored-by: Eden Reich <edenreich@users.noreply.github.com>