[Feature request] Add OpenVINO backend

[twoplan]

OpenVINO supports CPU, (i)GPU and NPU-enabled chips for Linux, Windows and macOS machines.

Could you please consider, if it's possible to add this library as a backend in Lc0 in order to run stuff on the NPU and GPU?

Comparing OpenVINO to Vulkan with llama.cpp and LLM Llama-3.2-1B-Instruct-f16.gguf shows OpenVINO@NPU equal to Vulkan@GPU, but with less power consumption. OpenVINO@GPU is more than twice as fast here.
Tested on Ubuntu 26.04 and Intel Core 258V laptop.

Thanks

OpenVINO NPU, GPU and CPU results
export GGML_OPENVINO_DEVICE=NPU
./llama-bench -m ~/models/Llama-3.2-1B-Instruct-f16.gguf -fa 1
OpenVINO: using device NPU

model	size	params	backend	ngl	fa	test	t/s
llama 1B F16	2.30 GiB	1.24 B	OPENVINO	99	1	pp512	2093.72 ± 18.93
llama 1B F16	2.30 GiB	1.24 B	OPENVINO	99	1	tg128	30.01 ± 0.11

export GGML_OPENVINO_DEVICE=GPU
./llama-bench -m ~/models/Llama-3.2-1B-Instruct-f16.gguf -fa 1
OpenVINO: using device GPU

model	size	params	backend	ngl	fa	test	t/s
llama 1B F16	2.30 GiB	1.24 B	OPENVINO	99	1	pp512	5748.75 ± 49.65
llama 1B F16	2.30 GiB	1.24 B	OPENVINO	99	1	tg128	27.33 ± 0.05

export GGML_OPENVINO_DEVICE=CPU
./llama-bench -m ~/models/Llama-3.2-1B-Instruct-f16.gguf -fa 1 --threads 4
OpenVINO: using device CPU

model	size	params	backend	ngl	fa	test	t/s
llama 1B F16	2.30 GiB	1.24 B	OPENVINO	99	1	pp512	210.14 ± 1.97
llama 1B F16	2.30 GiB	1.24 B	OPENVINO	99	1	tg128	12.92 ± 1.66

Vulkan GPU result
./llama-bench -m ~/models/Llama-3.2-1B-Instruct-f16.gguf -fa 1
load_backend: loaded RPC backend from ~/llama-b9553/libggml-rpc.so
ggml_vulkan: Found 1 Vulkan devices:
ggml_vulkan: 0 = Intel(R) Graphics (LNL) (Intel open-source Mesa driver) | uma: 1 | fp16: 1 | bf16: 1 | warp size: 32 | shared memory: 49152 | int dot: 1 | matrix cores: KHR_coopmat
load_backend: loaded Vulkan backend from ~/llama-b9553/libggml-vulkan.so
load_backend: loaded CPU backend from ~/llama-b9553/libggml-cpu-alderlake.so

model	size	params	backend	ngl	fa	test	t/s
llama 1B F16	2.30 GiB	1.24 B	Vulkan	-1	1	pp512	2073.07 ± 39.28
llama 1B F16	2.30 GiB	1.24 B	Vulkan	-1	1	tg128	38.80 ± 0.27