Introduce support for NF4 data type for OV weight compression

[l-bat]

What does this PR do?

• Added a new possible value for --weight-format CLI argument: nf4.
• Added a couple of tests for compression with the nf4 data type.

Usage examples:

Optimum CLI:

optimum-cli export openvino -m facebook/opt-125m --task text-generation-with-past --weight-format nf4 ./tmp

Optimum Python API:

model = OVModelForCausalLM.from_pretrained(
    model_id="gpt2",
    export=True,
    quantization_config=OVWeightQuantizationConfig(bits=4, weight_format="nf4"),
)

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?

[l-bat]

@AlexKoff88 please take a look

[nikita-savelyevv]

I believe AWQ and SE are not supported for NF4 on NNCF 2.13, so corresponding checks should be added. GPTQ is supported though. Possibly it could be combined with conditions for mxfp4 above to avoid code duplication.

[nikita-savelyevv]

Thanks!

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[AlexKoff88]

@l-bat, can you fix the style, please?