[CPU][Ref] Fix Reduce ops to produce stable (zero val) output for empty input

[mitruska]

Details:

• Fix Reduce ops to produce stable output (filled with zeroes instead of random values) for an empty input:
  including update of CPU and reference implementations.

• Such behavior has been requested for NNCF

• Originally random values has been produced, because common shape inference of the Reduce can produce non empty output shape for an empty input, for example Reduce with keep_dims=False:
  InputShape{2, 0}, reduce_axes=1, OutputShape{2}

• Common CPU Reduce::isExecutable was returning False for an empty input tensor (with 0 dim in the input Shape), and the Reduce::execute has not been executed at all for such case.
  The proposal of updated behavior is to return False for an empty output shape (with 0 dim in the output Shape),
  but when the input is empty, but the output is not, the output will be filled with 0.

Note: This change is considered as partially backward compatible with relatively low risk. Some of the Reduce ops already has defined default value other than zero in reference. Unifying the default value to be zero for all Reduce ops, makes onnx reduce prod and reduce min with empty input tests fail: logs

Alternatively Reduce ops common shape_inference can be updated:

• A. To produce empty output for empty input and keep 0 dim in the output shape even if requested by "axes" to be reduced.
  or
• B. Throw an error when reduction over 0 dimension is detected.

But those options are currently considered as not backward compatible, and would require new version of each Reduce* op.

Tickets:

• 117469