* add missing dict methods to DOK and tests

* turn off some dict dunders for dok_array

* tweak tests

* BUG: Optimize: NewtonCG min crashes with xtol=0

The minimize function crashes if `xtol` is set to `0`. This is
because the initial value of a loop variable is `2 * xtol`, which
is intended to ensure the loop runs at least once. However, if
`xtol = 0` then the loop never runs, a variable isn't instantiated
that is used outside of the loop, and an `UnboundLocalError` is
raised.

To fix this, the initial value of the loop variable is set to the
max float value.

* BUG: Optimize: UnboundLocalError for xtol=0 NCG

Corrected error from previous rework when switching to
`np.linalg.norm` that introduced the possibility of the unbound
variable error.

* TST: Optimize: NewtonCF Minimize when xtol=0

Regression test for scipygh-20214 which addressed an issue in the
Newton CG method which would raise an UnboundLocalVariable error
if `xtol` was set to zero. It also ensures the method finds the
solution in this case.

* MAINT: Add comment to explain initial value

Comment added to explain the purpose of setting `update_l1norm` to the max float value - this is to ensure the preceding while loop executes at least once.

Co-authored-by: Christian Lorentzen <lorentzen.ch@gmail.com>

* TST: Test solution value as well as success

This is for the `xtol=0` test in the Newton CG minimize method.

* TST: Explicitly extract val from array

---------

Co-authored-by: Christian Lorentzen <lorentzen.ch@gmail.com>

* related to scipygh-20300, though awaits `conda-forge` testing
on old MacOS versions for actual confirmation of fix

* pulls in changes described at mreineck/pocketfft#11
and scipy/pocketfft#2

* I'm honestly not sure how much we gain by being conservative and
not just turning the code path off entirely
(mreineck/pocketfft@33ae5dc);
perhaps there are some performance advantages or something, but if
we have any more trouble with this we should obviously just do that
(and I suspect that's what NumPy should do for
numpy/numpy#25940)

[skip cirrus] [skip circle]

[docs only]

See scipy#16930 for discussion.

closes scipygh-16930

This was introduced in scipygh-20193; no longer needed with pybind11 2.12.0

linalg.svd with too large matrices may segfault if
max(m, n)*max(m, n) > int_max on large-memory machines (on smaller memory machines
it may fail with a MemoryError instead). The root cause is an integer overflow
in indexing 2D arrays, deep in the LAPACK code.
Thus, detect a possible error condition and bail out early.

Co-authored-by: Pearu Peterson <pearu.peterson@gmail.com>

…e in `scalar_search_wolfe2`

* Converting amos to std::complex

* Define M_PI if not available (for windows)

* Bump test tolerance

---------

Co-authored-by: izaid <hi@irwinzaid.com>
Co-authored-by: Albert Steppi <albert.steppi@gmail.com>

* Fixes scipy#20294.

* Pull in new OpenBLAS binaries as discussed in the above issue,
to solve a bug that affects downstream consumers like `scikit-learn`.

* Some additional shims were needed: bumping `cibuildwheel` to
improve GitHub actions MacOS M1 runner support, and `openblas_support`
module needed an adjustment to process out an apparent syntax
error in the pkg-config file for openblas that we pull in from upstream.

…ate.RBFInterpolator` (scipy#20364)

* BUG: optimize: Fix wrong warning message if degree=-1 in interpolate.RBFInterpolator

* BUG: optimize: add a comment for test

* Prepare for SciPy `1.13.0` "final" release
instead of RC2 given urgency of supporting
NumPy 2 series.

* Remove note about SciPy `1.13.0` not being
released yet in the release notes, and update
the `1.13.0` release notes to reflect another
bump in the OpenBLAS version vendored in for PyPI
wheels.

* The usual version bumps removing `rc2` version
flags, and author/pr/issue list updates.