[FEATURE]: Export prompt tracing

**Context**

I'm investigating token usage differences between MCPJAM versions 0.6.0 and 0.7.0 to address the request in [#57](https://github.com/jeff-nasseri/mikrotik-mcp/issues/58). First, I want to say that the MCP tracing feature is excellent, and I'd like to suggest an enhancement.

**Request**

When tracing agent-to-MCP communication, I'd like to export and analyze token usage in detail. While the UI trace section shows tokens and the RAW view provides some information, I want to focus specifically on token consumption metrics.

Desired features:

1. Export token usage for individual steps (agent actions and MCP responses)
2. Categorize token usage per tool within a single prompt
3. Show token consumption for the LLM's tool-selection reasoning (before tool execution begins)
4. Additional token-related tracing insights as available

**Implementation Idea**

Similar to how browser DevTools lets you inspect network requests in detail, I'd like:

- Right-click context menu in the trace view to export a specific trace section
- Expandable details panel for each trace segment showing detailed token metrics
- A bird's-eye view of token distribution across the entire trace (timeline or breakdown by tool)

This would help optimize prompt engineering and understand where tokens are spent during agent execution.