Append basic geometry metrics to successful execute() responses

[pzfreo]

From a comparison with Casys-AI/mcp-build123d: their build123d_execute returns exact analytical metrics from the OCCT kernel on every run — volume, bounding box, surface area, and mass/center of gravity whenever a density is specified — without the client needing a second tool call.

Our execute() success path (tools/execute.py::execute_code) appends a Session objects: summary but no numbers; the assistant must follow up with measure() to learn anything quantitative. Since the recommended loop is "build one feature, then verify", nearly every execute is followed by a measure anyway — folding a lightweight metrics line into the execute envelope would save a round-trip per step and nudge assistants that skip verification into noticing when a boolean silently produced 0 mm³.

Suggested scope:

• On success, when current_shape is a 3D shape, append one line with bounding box and volume (e.g. current_shape: bbox 60.0 x 40.0 x 6.0 mm, volume 13,608 mm³); topology-light so it stays fast on heavy models
• Reuse the measure() machinery — measure(): add density parameter / material presets to report mass and physical inertia #237 already added density/material there; if a session-level density/material has been set, include mass too
• Keep it out of error responses (those already carry hints + failure classification)
• Possibly a CLI/env switch to disable it if the extra computation is unwanted on huge models