Fix Duplicate Eigenvalues in random_diagonalizable_matrix

[cxzhong]

random_diagonalizable_matrix accepted parallel lists of eigenvalues and
eigenspace dimensions. When the eigenvalue list contained duplicates, the
algorithm treated each occurrence as a separate eigenspace.

For example:

eigenvalues = [-5, 4, 4]
dimensions = [2, 3, 1]

was handled as three eigenspaces, even though the two entries for eigenvalue
4 should describe one eigenspace of dimension 4.

This caused the doctest failure under the specified random seed because the
randomly generated eigenvalues included a duplicate. The shifted matrix
B - 4*I could then have a reduced row echelon form with non-integral rational
entries, violating the documented teaching-oriented property of the constructor.

Fix

The algorithm now merges repeated eigenvalues before constructing the diagonal
matrix and eigenvector matrix.

The input:

eigenvalues = [-5, 4, 4]
dimensions = [2, 3, 1]

is normalized to:

eigenvalues = [-5, 4]
dimensions = [2, 4]

This matches the mathematical interpretation of eigenspaces for a
diagonalizable matrix.

📝 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

[github-actions]

Documentation preview for this PR (built with commit 3d87a02; changes) is ready! 🎉
This preview will update shortly after each push to this PR.