remove the abs() around the trapezoid segment sum so signed area comes through

The trapezoidal rule is a signed-area approximation: each linear segment
between (x1, f(x1)) and (x2, f(x2)) contributes (f(x1) + f(x2)) * (x2 - x1) / 2
to the running total. The previous code wrapped that in abs(), which forced
the running total to be non-negative. For functions that are positive in
the test range the abs() was a no-op and the existing doctests passed, but
two natural cases broke:

  - f(x) = -x**2 from -1.0 to 1.0 returns +0.6666... instead of -0.6666...
  - f(x) = x from -2.0 to 2.0 returns 4.0 instead of 0 (the symmetric
    linear function integrates to zero across the origin)

The doctests use f(x)=5 and f(x)=9*x**2 — both non-negative in the test
range, so they coincidentally worked. Drop the abs() and the formula now
returns the correct signed approximation for any real-valued f.

Verified with python3 -m doctest maths/area_under_curve.py: 5/5 pass.

Author: Zo Bot

maths/area_under_curve.py

@@ -41,7 +41,7 @@ def trapezoidal_area(
         # for trapezoidal area
         x2 = (x_end - x_start) / steps + x1
         fx2 = fnc(x2)
-        area += abs(fx2 + fx1) * (x2 - x1) / 2
+        area += (fx2 + fx1) * (x2 - x1) / 2
         # Increment step
         x1 = x2
         fx1 = fx2