feat(frontend): density contours and KDE for point layers

## Problem

Dense point layers look like solid blobs — users can't distinguish hotspots from uniform coverage, or outliers from noise.

## Proposal

Add kernel density estimation (KDE) + density contour rendering as a toggle on point layers, following [Geospatial Atlas](https://github.com/do-me/geospatial-atlas) / Embedding Atlas.

## Scope

- Frontend-only feature on point layers (vector MVT or WebGPU scatter)
- Toggle in layer controls: Points / Density / Both
- Adjustable bandwidth and opacity
- Order-independent transparency for overlapping points (from Embedding Atlas)

## Out of scope

- Server-side density rasters
- Heatmap on raster layers

## References

- https://apple.github.io/embedding-atlas

Start by locating the point-layer controls and the vector MVT or WebGPU scatter rendering entry points. Review the Geospatial Atlas and Embedding Atlas references, then define how the Points/Density/Both toggle, bandwidth, opacity, and order-independent transparency fit together. Done means point layers support the requested modes without adding server-side density rasters or raster-layer heatmaps.

Written by the indexing model from the issue text.