fix: show clear error when quickstart finds stale demo MySQL config

[lakshay-nasa]

Summary

• When a previous demo run left a MySQL DATABASE_URL in ~/.datahub/analytics-agent/.env, running quickstart would crash with a raw SQLAlchemy/pymysql traceback
• Now detects the connection failure and shows a clear, actionable message telling the user to reset their config directory

Test plan

• Run demo, then stop Docker, then run quickstart — verify the new error message appears
• Run quickstart with no prior config — verify normal flow is unaffected

[shirshanka]

Thanks for the fix @lakshay-nasa — accurate diagnosis of the stale DATABASE_URL problem.

I pushed a follow-up commit (5d8feaf) that promotes the message from "tell the user to rm -rf" to self-healing: when we see the demo's exact DATABASE_URL signature plus a connection error, we strip only that one line from .env and re-run bootstrap. The agent then comes up on the default SQLite DB and the user keeps their LLM API keys, conversation history, and any other config.

Two small helpers (_is_stale_demo_db_failure, _strip_env_vars) + 10 unit tests covering Mac/Linux URLs, the false-positive cases (user-supplied MySQL, unrelated stderr), and .env line-stripping edge cases.

Let me know if anything looks off — happy to iterate.