[SIP-215] MCP : Improve MCP chart preview tools - add rendered image and structured data preview options

Summary

Current Superset MCP service (6.1.0) exposes chart previews only in formats (ascii, url, table or vega lite) that does not provide exact chart previews as user will see in superset UI. We request first class MCP tools that return. -

A rendered images of the chart (faithful to Superset’s own rendering)
Structured chart data + stacking metadata sufficient to re render the chart in a downstream visualisation library
Use Case

we are integrating superset MCP into a conversational analytics assistant. When a user asks to preview or visualise a chart , we need to display it inline in the chat UI. Today this is not well supported.

Current Behaviour

get chart previews return previews as

ASCII
TABLES
Vega Lite
URL
These are not suitable for rendering a faithful, interactive chart in web ui.
Additionally, get chart info return only basic metadata and does not expose richer params/form data needed to reproduce the chart appearance.
Proposed Enhancements

Get chart preview - with format image - return a PNG/base 64 image of the actual rendered chart.
Get chart data enriched with richer metadata - include chart viz type, color scheme , axis config , metric/dimension labels and formatting so a downstream can render chart.
Why this matters

AI assistant and embedded analytics tools increasingly consume Superset via MCP. Faithful , programmatic chart previews are a common need. First class MCP support would make superset mcp significantly more useful for downstream agent/chatbot integrations.

Additional Context

Superset MCP version - 6:1:0
related : chart types like deck* cannot be re rendered by standard charting libraries at all making a server rendered image preview the only viable option for those.

Start at the Superset MCP service entry points for the current chart preview and chart info tools. Trace how the existing ASCII, table, Vega Lite, and URL responses are produced, then determine how rendered images and enriched chart metadata fit the API. Done means the tools expose faithful image previews and sufficient structured data and stacking metadata for downstream rendering.

Written by the indexing model from the issue text.