[Improve] Analyze performance of JPEG-compressed TIFF loading with nvJPEG+GDS

[gigony]

Loading JPEG-compressed TIFF image loading is done as part of cuCIM v22.02.00

If device parameter of read_region() method is "cuda", it loads a relevant portion of the image file (compressed tile data) into GPU memory using cuFile(GDS, GPUDirect Storage), then decompress those data using nvJPEG's Batched Image Decoding API.

However, current implementations are not efficient and performance is poor compared to CPU implementations.

We need to do a performance analysis of the implementation so that we can improve it.