change spatial_detection_score binning to be fixed bin size

[berl]

currently uses fixed number of bins for x and y, should be changed to fixed (square) bin size

[shachafl]

@berl looking to hear your thought before committing and PR:
I have replaced n_bins with pixel (LMK if a different name is better) where (pixel (int, optional): Pixel size in microns for spatial grouping. Defaults to 100.)
And also made the following addition and substitution of n_bins with nx_bins and ny_bins respectively:

    # determine number of bins on each axis for grouping the data spatially
    nx_bins = np.ceil((s2.x_centroid.max() - s2.x_centroid.min()) / pixel).astype(int)
    ny_bins = np.ceil((s2.y_centroid.max() - s2.y_centroid.min()) / pixel).astype(int)

    s2["xy_bucket"] = list(
        zip(
            pd.cut(s2.x_centroid, nx_bins, labels=list(range(nx_bins))),
            pd.cut(s2.y_centroid, ny_bins, labels=list(range(ny_bins))),
        )
    )

    ...
    bin_image_z_score = np.zeros([nx_bins, ny_bins])
    bin_image_difference = np.zeros([nx_bins, ny_bins])
    bin_image_ratio = np.zeros([nx_bins, ny_bins])
    bin_image_counts = np.zeros([nx_bins, ny_bins])

Also, using n_bins as a parameter for spatial_detection_scores() will return an error in my current code. But I can maintain the old parameter and functionality with an "if-else" statements if you prefer.

[berl]

sounds good!

1. I think bin_size will be less confusing with all the various image pixel sizes and conversions floating around in spatial users' minds
2. let's drop the n_bins . we can make sure the notebooks are updated