feat: map overhaul — pydeck heatmap, scatter, and 3D column layers

[jschloman]

Summary

Replace the current Swarm map with a multi-layer pydeck visualization using a CARTO dark basemap. Three composable layers toggled from the sidebar.

Depends on: #23 (theme foundation)

Basemap

No API key required. Use CARTO's free dark tile style:

CARTO_DARK = "https://basemaps.cartocdn.com/gl/dark-matter-gl-style/style.json"

Layer 1 — HeatmapLayer (density)

Show the geographic density of all check-ins as a smooth heatmap. Default layer when many check-ins exist.

pdk.Layer(
    "HeatmapLayer",
    data=checkins_df,
    get_position=["lng", "lat"],
    get_weight="visit_count",
    radius_pixels=60,
    intensity=1.2,
    threshold=0.25,
    color_range=[
        [12, 17, 32, 0],       # transparent (cold)
        [99, 102, 241, 120],   # indigo
        [34, 211, 238, 200],   # cyan
        [244, 114, 182, 255],  # pink (hot)
    ],
)

Layer 2 — ScatterplotLayer (individual venues)

Individual venue dots with tooltip showing venue name and visit count.

pdk.Layer(
    "ScatterplotLayer",
    data=venues_df,
    get_position=["lng", "lat"],
    get_color="[99, 102, 241, 200]",
    get_radius="radius",           # scaled by visit_count
    radius_min_pixels=4,
    radius_max_pixels=24,
    pickable=True,
)

Tooltip: {"text": "{place_name}\n{visit_count} visits"}

Layer 3 — ColumnLayer (3D visit frequency)

3D columns extruded by visit count, best viewed at 45° pitch. Toggle on from a sidebar checkbox.

pdk.Layer(
    "ColumnLayer",
    data=grid_df,              # venues aggregated to a ~100m grid
    get_position=["lng", "lat"],
    get_elevation="visit_count",
    elevation_scale=30,
    radius=80,
    get_fill_color="[99, 102, 241, 180]",
    pickable=True,
    auto_highlight=True,
)

Sidebar controls (new)

• Layer selector: st.radio("Map style", ["Heatmap", "Venues", "3D"])
• Pitch toggle: st.checkbox("3D view (45° tilt)") — sets ViewState(pitch=45) vs pitch=0

View state

Default: fit to bounding box of all check-ins. Compute center_lat, center_lng, and zoom from the data rather than hardcoding.

Acceptance criteria

• All three layers render correctly with CARTO dark basemap (no API key)
• Tooltip shows venue name and visit count on hover
• Sidebar controls switch between layers without full page reload
• View auto-fits to loaded check-in data
• Graceful empty state when no Swarm data loaded
• ruff check ., mypy, and pytest all pass