Fix inaccurate population choropleth description in NB04

The previous text claimed darker block groups cluster near the urban
core, but the map actually shows the opposite. Updated the description
to accurately reflect the pattern and explain why — smaller Census
block groups in denser areas — tying it back to the area-size bias
concept introduced earlier in the notebook.

Author: mihiarc

docs/notebooks/04-choropleth-maps.ipynb

@@ -206,11 +206,7 @@
   {
    "cell_type": "markdown",
    "metadata": {},
-   "source": [
-    "Notice how the map reveals spatial structure that would be impossible to see in a table. Darker block groups cluster near the urban core where apartment buildings concentrate residents, while lighter areas on the periphery tend to be lower-density single-family neighborhoods.\n",
-    "\n",
-    "Also note the legend on the right side -- SocialMapper automatically generates a color bar with the appropriate range and label."
-   ]
+   "source": "Notice how the map reveals spatial structure that would be impossible to see in a table. The downtown core and inner west side appear *lighter* (lower population per block group), while darker block groups are scattered across the east side and outer neighborhoods. This may seem counterintuitive -- isn't downtown denser? -- but remember that the Census Bureau draws **smaller block groups in denser areas** to keep each unit near 600–3,000 residents. A compact downtown block group and a sprawling suburban one can hold similar numbers of people, yet the suburban one appears larger and darker on the map.\n\nThis is exactly the **area-size bias** we mentioned earlier: raw population counts on a choropleth can mislead because block group *area* varies enormously. Normalized variables like median income and median age (which we map next) avoid this problem entirely."
   },
   {
    "cell_type": "markdown",