Added functions 'cudaSplit' and 'cudaMerge'

[HiroshiYAMA]

cudaSplit

This function splits from RGB/RGBA image to each color images, or splits to 3 color image and alpha channel image.
BGR/BGRA image is also okay.

cudaMerge

This function merges from each color images to RGB/RGBA image, or merges from 3 color image and alpha chennel image.
BGR/BGRA image is also okay.

Example1,
Python

# split.
img_IN = jetson.utils.loadImage('src.jpg', format='rgba8')
W = img_IN.width
H = img_IN.height
CH = img_IN.channels
img_SPLIT = [jetson.utils.cudaAllocMapped(width=W, height=H, format='gray8') for i in range(CH)]
jetson.utils.cudaSplit(img_IN, img_SPLIT)
jetson.utils.cudaDeviceSynchronize()

# merge.
img_OUT = jetson.utils.cudaAllocMapped(width=W, height=H, format='rgba8')
jetson.utils.cudaMerge(img_SPLIT, img_OUT)
jetson.utils.cudaDeviceSynchronize()

Example2,
Python

# split.
img_IN = jetson.utils.loadImage('src.jpg', format='rgba32f')
W = img_IN.width
H = img_IN.height
CH = img_IN.channels
img_COLOR = jetson.utils.cudaAllocMapped(width=W, height=H, format='rgb32f')
img_ALPHA = jetson.utils.cudaAllocMapped(width=W, height=H, format='gray32f')
jetson.utils.cudaSplit(img_IN, img_COLOR, img_ALPHA)
jetson.utils.cudaDeviceSynchronize()

# merge.
img_OUT = jetson.utils.cudaAllocMapped(width=W, height=H, format='rgba32f')
jetson.utils.cudaMerge(img_COLOR, img_OUT, img_ALPHA)
jetson.utils.cudaDeviceSynchronize()