PTQ for the OpenVINO backend

[andrey-churkin]

Changes

• The nncf.data.Dataset class was added. Also, available as nncf.Dataset. Please look at the documentation of this class to get more details.
• Added support of PTQ for the openvino.runtime.Model via Post-training Optimization Tool API. The method nncf.quantization.quantize() was added. Also, available as nncf.quantize(). This method applies post-training quantization to the provided model.
• The nncf.TargetDevice enum was added. It uses to specify the target_device parameter in the nncf.quantize() method.
• The nncf.QuantizationPreset enum was added. It uses to specify the preset parameter in the nncf.quantize() method.

Public API:

• nncf.Dataset
• nncf.quantize()
• nncf.TargetDevice
• nncf.QuantizationPreset

Reason for changes

The NNCF should support the PTQ for the OV model.

Related tickets

Ref: 90642

Tests