vulkan: dequantize iq4_xs 4 at a time

[netrunnereve]

I guess I missed this when I did the other quants 🤷‍♀️.

On my RX 470:

PR:

MUL_MAT(type_a=iq4_xs,type_b=f32,m=4096,n=512,k=14336,bs=[1,1],nr=[1,1],per=[0,1,2,3],k_v=0,o=1):                       42 runs - 24842.74 us/run -  60.13 GFLOP/run -   2.42 TFLOPS

model	size	params	backend	ngl	test	t/s
llama 1B IQ4_XS - 4.25 bpw	580.86 MiB	1.10 B	Vulkan	100	pp512	1116.81 ± 0.27

Master:

MUL_MAT(type_a=iq4_xs,type_b=f32,m=4096,n=512,k=14336,bs=[1,1],nr=[1,1],per=[0,1,2,3],k_v=0,o=1):                       38 runs - 26709.47 us/run -  60.13 GFLOP/run -   2.25 TFLOPS

model	size	params	backend	ngl	test	t/s
llama 1B IQ4_XS - 4.25 bpw	580.86 MiB	1.10 B	Vulkan	100	pp512	1055.85 ± 2.70

[0cc4m]

LGTM