Support models without a batch dimension

[Thakor-Yashpal]

Changes

• Modified nncf/quantization/algorithms/fast_bias_correction/openvino_backend.py.
• Updated the create_input_data method to add a specific check for 1D input shapes (i.e., len(shape) == 1).
• When a 1D model is detected, the function now bypasses the per-channel loop and directly uses the single mean tensor from the data list.

Reason for changes

The create_input_data method was looping based on the channel dimension size (3) but should have been based on the statistics list size (1)

Related tickets

Closes the feature request: "Currently FastBiasCorrection algorithm does not support models without the batch dimension" https://github.com/openvinotoolkit/nncf/issues/3481

Tests

Verified the fix by running the reproducer command provided in the issue, which now passes:

python -m pytest tests/openvino/native/test_fast_bias_correction.py -k OneDimMM

[MaximProshin]

@daniil-lyakhov , please take a look