Keep individual axis fits aligned with unsorted coordinates

[antonsoo]

Description

Baseline2D.individual_axes fits unsorted data against sorted coordinates, so coordinate-dependent methods can return the wrong baseline:

import numpy as np
from pybaselines import Baseline2D

x = np.array([2., 0., 3., 1.])
z = np.arange(5.)
data = x[:, None] ** 2 + z[None, :]
baseline, _ = Baseline2D(x, z).individual_axes(
    data, axes=0, method='poly', method_kwargs={'poly_order': 2})
print(np.allclose(baseline, data))
# Before: False
# After: True

Restore each axis's original coordinate order before constructing its one-dimensional fitter. That fitter then sorts the coordinates, data and optional weights together and returns results in the caller's order. No caller arrays are modified.

On three recorded-audio spectrograms, 90 of 144 row/column/permutation/method configurations disagreed with independently chained Baseline fits. All now match exactly, including intermediate baselines. All 36 sorted-coordinate controls retain identical output bytes. The regression covers row-only, column-only and combined permutations with exact quadratic surfaces.

Validation: the full suite passes (6967 passed, 113 skipped; baseline 6955 passed with the same skips and warnings), and ruff check . passes on Linux/Python 3.12. Other Python versions and platforms were not exercised locally.

Type of Pull Request

• Bug Fix
• New Feature
• Miscellaneous Changes (refactor, code improvements, etc.)
• Documentation or Example Programs

Pull Request Checklist

• New code and/or documentation is valid for use with the BSD 3-clause license.
• New code is fully documented with docstrings that follow Numpy style, if applicable.
• New code follows PEP 8 standards as closely as possible, if applicable.
• Added/updated tests and ensured they pass locally, if applicable.
• Verified that documentation builds locally, if applicable. No documentation changed.