Honor the step given in get_riemann_rectangles' x_range

[NgoQuocViet2001]

Motivation

get_riemann_rectangles accepts a three-element x_range so the caller can choose the sampling step — that is what the if len(x_range) < 3 branch normalises away. On that path two things go wrong in one line:

if dx is None:
    dx = self.x_range[2]
if len(x_range) < 3:
    x_range = [*x_range, dx]

rects = []
x_range[1] = x_range[1] + dx
xs = np.arange(*x_range)

np.arange steps by x_range[2], so reaching the requested right endpoint takes exactly one of those steps. The nudge uses dx, which on the three-element path was never set from x_range — it is self.x_range[2], the axis default.

The same line writes into the sequence the caller passed in. The two-element path builds a new list, so only the three-element path mutates the caller's object.

Measured on master with Axes() (default x_range (-8, 8, 1.0)):

axes = Axes()
graph = axes.get_graph(lambda x: x)
rects = axes.get_riemann_rectangles(graph, x_range=[0, 2, 0.5])

rectangles              : 5          (expected 4)
spans                   : (0.0,0.5) (0.5,1.0) (1.0,1.5) (1.5,2.0) (2.0,2.5)
spans past the right end: (2.0, 2.5)
caller's list afterward : [0, 3.0, 0.5]      <- was [0, 2, 0.5]

The last rectangle sits entirely outside the requested range, so the drawn area is wrong, and a second call with the same list drifts further because the bound has already been raised.

Proposed changes

• Copy x_range before writing to it.
• Nudge by x_range[2], the step actually being sampled.

Two-element x_range is byte-identical: dx is appended as x_range[2] there, so the two nudges are the same number.

Test

Two cases added to the existing tests/test_coordinate_systems.py — one asserting four rectangles ending at 2.0, one asserting the caller's list is unchanged.

python -m pytest tests/test_coordinate_systems.py tests/test_space_ops.py

Result: 8 passed. On master both new tests fail — five rectangles, and [0, 3.0, 0.5] where [0, 2, 0.5] is expected.