MCP integration — expose all databank columns (including custom ones) and add a bulk stats call

The MCP server in build 144 is genuinely useful, but get_strategy_stats only returns SQX's built-in statistics. Custom databank columns created via Custom Analysis snippets aren't available, and stats can only be fetched one strategy at a time. Both limits make it impractical to use the MCP connection for real portfolio work.

1. Return custom databank columns (highest value)

Many users add custom scoring columns via Custom Analysis — robustness scores, retention metrics, composite rankings. These are often the columns we actually select strategies on, but they're invisible over MCP, so any AI-assisted analysis has to fall back to manual CSV exports.

Ideally get_strategy_stats would return whatever columns the databank's currently selected view displays, so custom and built-in columns come through identically and the view acts as the column filter.

2. Add a bulk endpoint

Fetching stats one strategy at a time doesn't scale — a databank of 250 strategies means 250 round trips, and calls time out on larger projects. A single call such as get_databank_stats(project, databank) returning all rows at once (or paginated) would make analysis across a full databank practical.

3. Optional: expose the export function

An MCP equivalent of "Export databank contents" — returning the databank as CSV — would achieve points 1 and 2 in one step and require little new logic, since the export already exists in the UI.

Why this matters

The current workflow for AI-assisted analysis is: manually export each databank to CSV, sync the files somewhere the assistant can read them, repeat every time anything changes. With the above, the MCP connection would replace that entirely. Given build 144 positions SQX as an AI-integrated platform, closing this gap makes the integration genuinely usable rather than a demonstration.