Update INC 3.0 New API for Keras Static Quant

[zehao-intel]

Description

This PR includes the code of the implementation of INC 3.0 New API for Keras Static Quant.

• quantize_model API aligned with TFMOT.
• Support Static Quant for Keras model

More algorithms might be added in later PRs.

3.0 API Usage

Users should pass the model, quant_config and calib_dataloader as arguments to this quantize_model API.
model: A tf.keras.Model instance
quant_config: A KerasStaticQuantConfig instance, the usage is the same as nc.config.PostTrainingQuantConfig.
calib_dataloader: A Dataloader for calibration.

PTQ for Keras models

from neural_compressor.tensorflow import StaticQuantConfig, quantize_model

config = StaticQuantConfig(backend="itex")
model = keras.models.load_model("./baseline_model")
q_model = quantize_model(model, config, calib_dataloader=CustomDataloader())

How has this PR been tested?

Pre-CI

Dependency Change?

No

[yiliu30]

The 3.x tf binary includes neural_compressor.common and neural_compressor.tensorflow instead of neural_compressor.keras. Please align with @chensuyue.

[chensuyue]

Please move 3.x test here: https://github.com/intel/neural-compressor/tree/master/test/3x
I will add separate CI for keras 3.x.

[zehao-intel]

All comments above have been solved.