- Create 2.4.6-notes.rst
- Update release.rst
- Update pyproject.toml

[skip CI]

MAINT: Prepare 2.4.x for further development

I'll make these helpers static-inline functions on main.
Unfortunately, this doesn't just cause wrong promotion for bool->uint8
in `arr.conj()`, but it also leads to all other non-complex types
returning unnecessary copies which seems like a pretty unfortunate
regression, even if it might not be very visible in practice.

I am suspecting we may need a 2.4.6 for this :(

BUG: Fix regression in ``arr.conj()``

…31347)

* Explicitly remove 0 from eigenvalue signs to ensure that vh is unitary.
* Add Hermitian SVD test that has 0 as a singular value.

…ator. (#31456)

…ator. (#31456) (#31460)

Co-authored-by: Warren Weckesser <warren.weckesser@gmail.com>

BUG: `np.linalg.svd(..., hermitian=True)` returns non-unitary `vh` (#31347)

- Create 2.4.6-changelog.rst
- Update 2.4.5-notes.rst

[skip actions] [skip azp]

REL: Prepare for the NumPy 2.4.6 release