Fix: test for windowed projection in analysis.Video class

[janreising]

section of get_image_project()

...

from numpy.lib.stride_tricks import sliding_window_view

Z, X, Y = img.shape
proj = np.zeros((X, Y))

z_step = int(window/2)
for x in tqdm(range(X)) if show_progress else range(X):
    for y in range(Y):

        slide_ = sliding_window_view(img[:, x, y], axis=0, window_shape = window) # sliding trick
        slide_ = slide_[::z_step, :] # skip most steps
        agg = agg_func(slide_, axis=1) # aggregate
        proj[x, y] = window_agg(agg) # window aggregate

Current implementation of test:

proj_org = np.zeros(original_data.shape)
                    for x in range(original_data.shape[1]):
                        for y in range(original_data.shape[2]):
                            proj_org[:, x, y] = pd.Series(original_data[:, x, y]).rolling(window=window).apply(proj_func).values

This fails because the sliding trick windows differently than intuitively expected. Eg. initial values of sliding trick produces np.nan