This fixes a bug with rounding pixels to the nearest palette color. Specifically, there is a typo with computing the minimum distance from a palette color to a given cache box. This is causing palette colors to be incorrectly excluded from consideration.

The bug can be observed in the following example:

from PIL import Image
palette = Image.new('P', (1, 1))
palette.putpalette([0, 0, 0, 32, 32, 32] + ([0, 0, 0] * 254))
image = Image.new('L', (9, 1))
image.putdata([0, 4, 8, 12, 16, 20, 24, 28, 32])
quantized_image = image.convert('RGB').quantize(
    dither=Image.Dither.NONE, palette=palette)
print(list(quantized_image.convert('L').getdata()))

Before the change, this prints [0, 0, 0, 0, 0, 0, 0, 0, 32]. After the change, it prints [0, 0, 0, 0, 0, 32, 32, 32, 32].

Changes proposed in this pull request:

• Fix bug with rounding pixels to palette.
• Update tests to account for fix.