Improve dashboard and card parameter workflows

[andreagrandi]

Summary

Closes #14. Running parameterized cards and dashboard cards previously required knowing parameter IDs and values up front, with no way to discover them. This adds discovery commands and makes failed runs point at the exact command to inspect valid parameters.

Changes

• card params <id> — lists the parameters a saved question accepts (from native template tags). card and snippet template tags are excluded since they reference other queries rather than taking a user value. Table output ends with a ready-to-edit card run ... --param name=<value> line.
• dashboard params list <dashboard-id> — lists a dashboard's parameters, types, and how many cards each filters; completes the dashboard params group (list / values / search).
• Actionable run errors — a failed parameterized run now surfaces a ParameterizedQueryError carrying the precise discovery command (mb-cli card params <id> / mb-cli dashboard params list <id>), shown in both text and --error-format json output.
• Metabase rejects invalid card parameters with HTTP 500 (not 400), so the error wrapping keys off whether the caller supplied --param flags rather than the status code.
• Updates the embedded agent context with the new commands, a parameterized-workflow section, and examples.

Testing

• make fmt, make vet, make build, make test all clean
• New tests cover parameter discovery (cards with/without parameters, dashboards) and invalid-parameter paths for both card run and dashboard run-card
• Smoke tested against a live Metabase instance: card params, card run (valid + invalid params), dashboard params list, dashboard run-card invalid-param path

Changelog

Entry added under ## [Unreleased].