Added ImageSequence all_frames

[radarhere]

Resolves #3758

Adds a new method to apply a supplied function to all frames of an image at once. So, for example,

from PIL import Image, ImageSequence
im = Image.open('Tests/images/g4-multi.tiff')

ims = ImageSequence.all_frames(im, lambda im_frame: im_frame.rotate(180))

ims[0].save('out.tiff', save_all=True, append_images=ims[1:])

[pirate486743186]

hmmm.... Maybe add multi-threading support?
The function could take a third parameter, the maximum number of threads to be used, default to one.

If no time for this, then you could just silently add the third parameter anyway, so that future code is backwards compatible with the first version of the function.

[radarhere]

While we are concerned with backwards compatibility, I don't think that we are concerned with forwards compatibility.

Nothing else in the Pillow runtime uses multiprocessing, so I'm not personally inclined to integrate it here. If anyone else feels otherwise, feel free to comment. Multiprocessing can't be applied to the process of splitting up a single image into frames, because it uses the same file pointer, but it could be applied to the supplied function. If a user wanted to do that, they could just -

from PIL import Image, ImageSequence
im = Image.open('Tests/images/iss634.gif')

ims = ImageSequence.all_frames(im)

from multiprocessing import Pool
def f(im_frame):
    return im_frame.rotate(180)
ims = Pool(5).map(f, ims)

ims[0].save('out.gif', save_all=True, append_images=ims[1:])

[pirate486743186]

but isn't this a special case? You are basically processing multiple images at once. And it seems it would be used frequently IMHO.

[hugovk]

Please add release notes.