Make unwrap_phase fail for 2D input containing NaNs

[lagru]

Description

Closes #3831.

After some stone-age debugging with print statements, it seems to me that the infinite loop happens inside

scikit-image/skimage/restoration/unwrap_2d_ljmu.c

Lines 128 to 139 in 97b3ecb

	EDGE *partition(EDGE *left, EDGE *right, double pivot) {
	while (left <= right) {
	while (left->reliab < pivot) ++left;
	while (right->reliab >= pivot) --right;
	if (left < right) {
	swap(*left, *right);
	++left;
	--right;
	}
	}
	return left;
	}

which never exits due to pivot being NaN. Current solution is to check pivot before passing it to that function and return error codes up the call stack until we can raise a proper Python exception.

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

Summarize the introduced changes in the code block below in one or a few sentences. The
summary will be included in the next release notes automatically:

Raise a `RuntimeError` in `skimage.restoration.unwrap_phase` If a NaN is encountered 
while sorting edges. This may happen if the input `image` contains unmasked NaNs.

[lagru]

Not sure if that is the best way to address the problem. I understand little of what's going on, but it's better than hanging with one core stuck at 100%. Performance wise I didn't see any significant impact for the following snippet taken from plot_phase_unwrap.py:

import numpy as np
from skimage import data, img_as_float, color, exposure
from skimage.restoration import unwrap_phase
image = color.rgb2gray(img_as_float(data.chelsea()))
image = exposure.rescale_intensity(image, out_range=(0, 4 * np.pi))
image_wrapped = np.angle(np.exp(1j * image))

%timeit unwrap_phase(image_wrapped)

Though I am a bit dubious because the paper makes it seem like there should be a huge amount of edges and I am not sure how often quicker_sort is called in the total scheme of things.

[lagru]

Ugh, this snippet (from #3831 (comment)) still fails

import numpy as np
from skimage.restoration import unwrap_phase
wrapped = np.empty((100, 100), dtype=np.float64)
wrapped[:] = np.linspace(0, 4*np.pi, 100) % (2*np.pi)
wrapped[10, 10] = np.nan  #  <-- comment out to get working code
masked = np.ma.masked_array(wrapped, mask=np.isnan(wrapped))
print(np.isnan(masked).any())
unwrapped = unwrap_phase(masked)

[github-actions]

Hello scikit-image core devs! There hasn't been any activity on this PR for more than 180 days. I have marked it as "dormant" to make it easy to find.
To our contributors, thank you for your contribution and apologies if this contribution fell through the cracks! Hopefully this ping will help bring some fresh attention to the review. If you need help, you can always reach out on our forum
If you think that this PR is no longer relevant, you may close it, or we may do it at some point (either way, it will be done manually). If you think the PR is valuable but you no longer have the bandwidth to update it, please let us know, so that someone can take it over. 🙏