test: make assert_linequal ignore dimension order

[FBumann]

I hit some false positives while working on sth else. This fixes them. Making the testing more robust.

Note

The following content was generated by AI.

What this does

linopy expressions are dimension-labelled xarray structures, so they inherit
xarray's broadcast order-sensitivity: x + y produces dims (x_dim, y_dim)
while y + x produces (y_dim, x_dim). That difference is not mathematically
meaningful, but xarray's assert_equal is order-sensitive, so
assert_linequal(x + y, y + x) failed.

assert_linequal already normalizes term order before comparing (its
docstring promises "semantically equal"). This aligns dimension order the
same way: a new _align_dim_order helper transposes the second operand to the
first's dimension order when they share a dimension set. If the dimension sets
differ, the expressions are genuinely unequal and the comparison is left
untouched so assert_equal reports the real difference.

Notes

• Touches only linopy/testing.py; adds test/test_testing.py (dim-order
  insensitivity, plus genuinely-different expressions still fail).
• Full test_linear_expression.py + test_quadratic_expression.py pass
  unchanged (333), i.e. nothing was relying on the order-sensitive behavior.

Reproducer

import pandas as pd
from linopy import Model
from linopy.testing import assert_linequal

m = Model()
a = m.add_variables(coords=[pd.Index([0, 1], name="i")], name="a")
b = m.add_variables(coords=[pd.Index([0, 1, 2], name="j")], name="b")

(a + b).data.coeffs.dims   # ('i', 'j', '_term')
(b + a).data.coeffs.dims   # ('j', 'i', '_term')
assert_linequal(a + b, b + a)   # before: AssertionError; after: passes

[codspeed-hq]

Merging this PR will not alter performance

✅ 138 untouched benchmarks
⏩ 138 skipped benchmarks¹

---

_{Comparing fluxopt:fix/assert-linequal-dim-order (3d33893) with master (fe798b1)}

Footnotes

1. 138 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[FBumann]

canceled pypsa-tests