Fix constrained minimization dispatch errors

[cxzhong]

Fixes #42711.

The dispatch logic in minimize_constrained previously selected TNC for every
algorithm name other than the exact string l-bfgs-b, indexed an empty
constraint list, and left the result uninitialized for unsupported constraint
shapes and ordinary callable objects.

This pull request:

• validates the two documented algorithm names and reports unknown values;
• treats an empty constraint list as unconstrained bounds for TNC and L-BFGS-B;
• accepts partial functions, bound methods, callable instances, and builtins;
• reports invalid constraint specifications with a descriptive TypeError;
• preserves the historical symbolic-constraint representation and bounds
  dispatch compatibility.

The doctests cover both bound solvers, symbolic and Python objectives,
arbitrary callable constraints, invalid inputs, and heterogeneous interval
sequences accepted by SciPy.

📝 Checklist

• The title is concise and informative.
• The description explains in detail what this PR is about.
• I have linked a relevant issue or discussion.
• I have created tests covering the changes.
• I have updated the documentation and checked the documentation preview.

⌛ Dependencies

None.

[github-actions]

Documentation preview for this PR (built with commit 32e36ef; changes) is ready! 🎉
This preview will update shortly after each push to this PR.