Enable autoround quantization

[echarlaix]

Needs neural-compressor >= v2.6.0 to avoid incompatibility with itrex

coming from missing n_samples argument and train_bs / bs mismatch : https://github.com/intel/intel-extension-for-transformers/blob/v1.4.2/intel_extension_for_transformers/transformers/llm/quantization/utils.py#L397 https://github.com/intel/neural-compressor/blob/v2.5.1/neural_compressor/adaptor/torch_utils/auto_round.py#L19

[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.

[echarlaix]

ITREX was deprecated in : https://github.com/huggingface/optimum-intel/pull/880