Exercise 2: Uppercase Measures Wider Than Lowercase — Possible Solution
====================================================================

THE TEST
------------------------------
    char_width_px('a', 16.0)
    char_width_px('A', 16.0)

RESULT
------------------------------
    'a' -> 8.0     (CHAR_WIDTHS_EM['a'] = 0.50, times 16.0)
    'A' -> 10.72   (CHAR_WIDTHS_EM['A'] = 0.67, times 16.0)

'A' is genuinely wider than 'a' at the identical 16px font-size --
10.72 > 8.0.

WHY THE TABLE ASSIGNS DIFFERENT EM VALUES TO THE SAME LETTER
------------------------------
CHAR_WIDTHS_EM stores 'a' and 'A' as two completely separate dictionary
keys, each with its own independently-chosen em value (0.50 for the
lowercase form, 0.67 for the uppercase form). There's no shared
"letter A" concept in the table at all -- case is simply part of the
character's own identity for lookup purposes, exactly the same way a
real font's own glyph outlines for lowercase and uppercase letters are
two entirely separate drawings, not a single shape scaled or reused.

WHY THIS MATCHES REAL TYPOGRAPHY, NOT AN ARBITRARY CHOICE
------------------------------
In the overwhelming majority of real fonts, capital letters are drawn
with more strokes, wider bowls, and generally more visual "bulk" per
character than their lowercase counterparts -- an 'A' has two diagonal
strokes plus a crossbar; a lowercase 'a' is a single compact,
rounder shape. This isn't specific to any one font; it's a near-
universal property of Latin-script typography, which is exactly why a
plausible, hand-estimated width table (like this chapter's own
simplified one) can assign uppercase letters systematically larger em
values across the board -- most of the uppercase entries in
CHAR_WIDTHS_EM (0.61-0.94) sit meaningfully higher than their
lowercase counterparts (0.28-0.78), reflecting this same real-world
pattern consistently, not just for 'A' specifically.

WHY THIS WORKS AS AN ANSWER
------------------------------
This exercise demonstrates the table captures a genuine STRUCTURAL
property of real text (case affects width, systematically, not
randomly) rather than being an arbitrary or decorative set of numbers.
It also reinforces why a flat, single-width-per-character-count model
(this chapter's own predecessor bug) can never be fixed by simply
picking "one better number" -- the actual distribution of character
widths in real text depends on which SPECIFIC characters appear, case
included, not just how many characters there are.