GeezerVison

• I hereby agree to dual licence this and any previous contributions under both
  the Biopython License Agreement AND the BSD 3-Clause License.

• I have read the CONTRIBUTING.rst file, have run pre-commit
  locally, and understand that continuous integration checks will be used to
  confirm the Biopython unit tests and style checks pass with these changes.

• I have added my name to the alphabetical contributors listings in the files
  NEWS.rst and CONTRIB.rst as part of this pull request, am listed
  already, or do not wish to be listed. (This acknowledgement is optional.)

Closes #5342

Zero-length vectors currently produce NaN when calculating the cosine in Vector.angle,
but the roundoff clamp converts that NaN into -1. As a result, four collinear
points return a finite dihedral of -pi, although the angle is undefined.

This change raises ValueError for zero-length vector angles and uses a
NaN-preserving clamp for roundoff. It also uses the scalar triple product to
determine the dihedral sign, so valid planar dihedrals do not require an angle
with a zero intermediate vector. The 180-degree endpoint is returned as +pi,
consistent with the documented interval. vector_to_axis uses a dot-product
projection so projecting the origin remains valid.

Seven new test methods cover zero vectors in either argument, NaN propagation,
roundoff, coincident points, undefined planes, signed and planar dihedrals, and
projection of the origin. The API docstrings and release notes document the
new exception behavior.

Behavior to review

This proposes ValueError for undefined zero-length angles rather than returning
NaN with a warning. calc_angle and calc_dihedral inherit that behavior.
Callers that previously accepted a finite result for degenerate geometry may now
need to handle the exception. Feedback on this API choice is welcome.

Validation

• Python 3.12.14, NumPy 2.3.5, Windows 11.
• 31 focused vector, polypeptide and exposure tests passed. The added regressions
  fail against unmodified Biopython 1.88.
• The offline Bio.PDB run passed 26 module groups and skipped three:
  test_PDB_alphafold_db, test_PDB_PDBList (offline), and
  test_PDB_PSEA (external executable unavailable).
• 1,000 seeded nondegenerate dihedrals matched an independent atan2 calculation
  to a maximum absolute difference of approximately 5.1e-14 radians.
• All applicable configured pre-commit checks passed for the three changed files.
• The offline full runner completed and reported 502 test entries, with two
  failing module groups: test_Align_aligncore and
  test_align_substitution_matrices. The failing checks also reproduce against
  the unmodified 1.88 wheel.

The local environment lacks a source-build C toolchain. Validation used the
current Python source with the official 1.88 wheel's extension binaries in a
separate staging directory. The two failing groups exercise newer C-extension
behavior that those binaries do not provide. This is not a clean source-build
pass; source-built CI validation is still needed. Online and unavailable
external-tool tests were skipped.

This contribution was prepared with AI assistance; the implementation, regression
results, and validation limitations are documented for review.