FIX: Keep adaptive-minmax constraints aligned with unsorted coordinates

[antonsoo]

Description

adaptive_minmax builds endpoint constraints in sorted coordinate order. When weights is omitted, it skips restoring those constraints to the input order, so reversed or shuffled coordinates constrain the wrong samples. This affects both Baseline and Baseline2D.

Always restore the constrained weights to input order. Keep the existing conditional restoration for the original weights. Add regressions for both polynomial methods and document that endpoints follow physical coordinates.

Validation:

• The eight new reversed/shuffled cases fail before and pass afterward; the four ordered controls pass throughout. The full suite passes with 10,024 passed and 113 skipped. Ruff and the full documentation build, including all 23 gallery examples, pass.
• On measured Raman spectra, all 144 incorrect reordered results are corrected: 96 single-spectrum fits and 48 fits of matrices containing repeated measurements. All 216 tested outputs now exactly match coordinate-aligned references, including an independent one-dimensional oracle that constructs endpoint membership from physical coordinates. All 360 explicit-weight control outputs remain byte-identical.

These checks establish coordinate invariance; they do not assess the quality of the baseline model.

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 follows the style and docstring guidelines
• Added/updated tests and ensured they pass locally
• Verified that documentation builds locally

[derb12]

The segmentation faults in the test-min-dependencies job are unrelated to the changes here, so can ignore.

Thanks again for both this and PR 97.