Auto-promote more keyword arguments for trust-region algorithms

[jessegrabowski]

Several arguments for trust-region minimization algorithms (trust-ncg, trust-krylov, trust-exact) are not documented. This PR moves all valid kwargs into the options dictionary for users.

[Copilot]

Pull Request Overview

This PR improves the documentation and auto-promotion of keyword arguments for trust-region minimization algorithms (trust-ncg, trust-krylov, trust-exact) by moving previously undocumented but valid kwargs into the options dictionary for better user accessibility.

• Expands the valid_options lists for trust-region algorithms to include previously undocumented parameters
• Updates test utilities to handle undocumented SciPy options by maintaining a hardcoded mapping
• Ensures better coverage of available options for users of trust-region optimization methods

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
better_optimize/constants.py	Adds previously undocumented options like "workers", "subproblem_maxiter", "initial_trust_radius", "max_trust_radius", "eta", and "gtol" to the valid_options for trust-region algorithms
tests/test_utilities.py	Introduces hardcoded mapping of undocumented options per algorithm and updates assertion logic to account for these missing documentation gaps

Comments suppressed due to low confidence (1)

tests/test_utilities.py:184

• The error message could be more descriptive. Consider including the method name in the error message: f'Missing options for {method}: {", ".join(missing_options)}'

    assert not missing_options, "missing options: " + ", ".join(missing_options)