docs(telemetry): document summary.tokens.llm.prompt_cached and completion_reasoning (#2305)

[r266-tech]

What

#2305 started serializing two new LLM token counters in the operation-telemetry summary, but the telemetry guide still lists only input / output / total. This adds the two missing fields to the EN + ZH guides so developers can discover them.

Drift

In openviking/telemetry/operation.py, OperationTelemetry.build() now emits:

"llm": {
    "input": llm_input_tokens,
    "output": llm_output_tokens,
    "total": llm_total_tokens,
    "prompt_cached": llm_prompt_cached_tokens,            # tokens.llm.prompt_cached
    "completion_reasoning": llm_completion_reasoning_tokens,  # tokens.llm.completion_reasoning
},

These are sourced from provider usage details (add_token_usage_by_source(..., prompt_cached_tokens, completion_reasoning_tokens), recorded only when source == "llm"; see models/vlm/base.py docstring: "Number of cached prompt tokens / reasoning completion tokens from provider usage details"). But docs/{en,zh}/guides/07-operation-telemetry.md documented neither field, so anyone reading the telemetry summary to attribute LLM cost (cache savings / reasoning-token spend) couldn't find them.

Change (docs only, EN + ZH mirror)

• Add summary.tokens.llm.prompt_cached and summary.tokens.llm.completion_reasoning rows to the summary.tokens field table in both guides.
• Add the two keys to the JSON example's llm block.
• Example values are non-zero (prompt_cached: 8, completion_reasoning: 4) on purpose — the guide states "numeric 0 values are omitted from the response", so a 0-valued example would contradict that contract.

No code or behavior change.

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

🎫 Ticket compliance analysis 🔶 2305 - Partially compliant Compliant requirements: (Docs only) Documents the new prompt_cached and completion_reasoning token counters added in feat(eval)Opt vikingbot eval script #2305 Non-compliant requirements: (No code changes in this PR; focuses solely on documentation) Requires further human verification: All requirements (except docs) are unrelated to this PR and require verification from the original feat(eval)Opt vikingbot eval script #2305 code changes

⏱️ Estimated effort to review: 1 🔵⚪⚪⚪⚪

🏅 Score: 100

🧪 No relevant tests

🔒 No security concerns identified

✅ No TODO sections

🔀 No multiple PR themes

⚡ No major issues detected

[github-actions]

PR Code Suggestions ✨

No code suggestions found for the PR.