Update video duplicate finder and more

[qarmin]

• updated vid_dup_lib_finder to latest version
• used jxl -> image-rs converter provided by jxl library
• using bigger buffer size to speedup checking for files in HDD and also SSD(biggest gains are in HDD)
• option to enable fast-image-resize

Performance comparison of using Array and Vector with specific buffer sizes for reading files from disk and calculating their hashes.
This is quite realistic scenario which also uses rayon which sometimes sometimes mess with predictability of results.
My computer have quite good CPU, but cheap Sata SSD, so results shows disk

(Time to read files and calculate hashes in parallel, smaller is better)

Name	250000 files ~50 KB(SSD)	170 files 5MB-150MB(SSD)	1 file 0.9 GB(SSD)	6200 files 50KB - 50MB(HDD)	1 file 671 MB(HDD)
Array 16KB	Base	Base	Base	Base	Base
Vector 16KB	0%	0%	0%	0%	0%
Vector 1MB	-7%	-4%	-16%	-45%	0%
Thread local Vector 1MB	-12%	-4%	-16%	-45%	0%

I tried to use locks to read at max 1 file from hdd, but there was no performance gains

There is also option to speedup resizing images, which causes significant speedup when checking for similar images
On my OS and hashed 91 files(~3/4 MB each, 3000x4000 JPGS) I got sometimes even 3x speedup(in real world, speedup should be smaller, depending on size of files(bigger should get bigger gains) and algorithm(nearest and blockhash almost not have any speedup because one is very simple to resize and other do not resize image before hashing ))

Size	Algorithm	Filter	NORMAL [ms]	FAST IMAGE RESIZE [ms]
16	Mean	Gaussian	8845	2738
16	Gradient	Nearest	2127	2216
16	Gradient	Lanczos3	8802	2809
8	Gradient	Lanczos3	8819	2771
64	Gradient	Lanczos3	5347	2486
16	BlockHash	Lanczos3	6162	5934