Fix shape and reformat free tensor handling in the input byte size check

[pskiran1]

• Added support for the is_non_linear_format_io flag in the model configuration to identify whether input/output tensors use a non-linear IO format.
• Modified the auto-complete configuration to correctly initialize the is_shape_tensor and is_non_linear_format_io flags.
• For tensors using the non-linear IO format (reformat-free tensors), the backend now validates the input byte size.

[pskiran1]

@tanmayv25, Thank you for the approval. I added a new commit 5b8c263 to improve the error message to match the remaining tensor types error message in the core library. Could you please re-approve? Thank you.