fix(pipeline): ground local-agent system prompt with current date

[leonoxo]

Summary

• The local-agent runner's system prompt is a static string with no template-variable support, so the model has no anchor for "today" and resolves relative time references (e.g. "this quarter", "latest", "currently") against whichever period is best represented in its training data instead of the actual current date — this can produce confident, stale answers to time-sensitive questions (e.g. answering a "this quarter's revenue" question with year-old earnings data) instead of triggering a search tool.
• PreProcessor now appends a short, deterministically-computed Current date: YYYY-MM-DD (Weekday) note (plus guidance to verify time-sensitive facts with a search tool rather than answering from memory) to the system prompt on every request for local-agent pipelines — this is independent of LLM tool-call judgment, unlike binding a "get current time" MCP/tool, which the model may or may not choose to call.
• Refactored the existing skill-awareness prompt injection to share the same _append_to_system_prompt helper instead of duplicating the append-to-system-message logic.

Test plan

• uv run pytest tests/unit_tests/pipeline/test_preproc.py -q — 13 passed (10 existing + 3 new covering: append to existing system message, create system message when none exists, skip for non-local-agent runners)
• uv run pytest tests/unit_tests -q — 2664 passed, 1 skipped
• uv run ruff check / ruff format --check on changed files

[github-actions]

All contributors have signed the CLA. ✅ 所有贡献者均已签署 CLA。
_{Posted by the CLA Assistant Lite bot.}

[leonoxo]

I have read the CLA Document and I hereby sign the CLA