Set LDFLAGS and CPPFLAGS for keg-only libomp in CI config

[lagru]

Description

Closes #7406. The output while running "brew install libomp" suggests that the mentioned flags need to be set accordingly. So check whether this addresses the issue.

Checklist

• A descriptive but concise pull request title
• Docstrings for all functions
• Unit tests
• A gallery example in ./doc/examples for new features
• Contribution guide is followed

Release note

For maintainers and optionally contributors, please refer to the instructions on how to document this PR for the release notes.

...

[lagru]

Great news the compilation works now. However, the tests fail...

=========================== short test summary info ============================
FAILED measure/_moments.py::skimage.measure._moments.moments_hu
FAILED measure/tests/test_fit.py::test_ellipse_parameter_stability - AssertionError: 
Arrays are not almost equal to 7 decimals
 ACTUAL: 2.356194490192345
 DESIRED: 0.7853981633974483
====== 2 failed, 9088 passed, 3 skipped, 12 warnings in 221.30s (0:03:41) ======

[lagru]

@stefanv do you maybe want to take a look at the failing test_ellipse_parameter_stability?

[jarrodmillman]

Note that macos-latest switched from 12.7.4 to 14.4.1.

[stefanv]

The moments failure is innocuous:

Expected:
    array([0.74537037, 0.35116598, 0.10404918, 0.04064421, 0.00264312,
           0.02408546, 0.        ])
Got:
    array([ 7.45370370e-01,  3.51165981e-01,  1.04049179e-01,  4.06442107e-02,
            2.64312299e-03,  2.40854582e-02, -8.90690725e-21])

[stefanv]

The ellipse failure: the ellipse parameters a and b are correctly estimated (good!). The angle is given as 135 instead of 45. I don't think a pi/2 correction is needed in this case, yet it was added.

        # angle of counterclockwise rotation of major-axis of ellipse
        # to x-axis [eqn. 23] from [2].
        phi = 0.5 * np.arctan((2.0 * b) / (a - c))
        if a > c:
            phi += 0.5 * np.pi

a is definitely larger than c in this case, but what does that comparison do?

The code above is from #2482

I think we should carefully look at https://mathworld.wolfram.com/Ellipse.html#eqn23 too. Question is: why does this pop up now, and only on newer macos / clang compiler? They must have a math library with different trigonometric branching.

The parameter stabilization, just below the above, seems fine:

        # stabilize parameters:
        # sometimes small fluctuations in data can cause
        # height and width to swap
        if width < height:
            width, height = height, width
            phi += np.pi / 2

[jarrodmillman]

As a temporary workaround, we can merge #7408. This PR will need to be rebased and revert that commit. Also note that macos-latest is now macos-14-arm64 (it used to be macos-12):