Fix overflow in collectors

[AlexanderDokuchaev]

Changes

Cast float16 and float32 to float64 for mean and median.

Reason for changes

Error on calculation parameters of FQ for large scale.

Tests

test_reducers_and_aggregators.py::test_overflow

[codecov]

Codecov Report

Attention: 31 lines in your changes are missing coverage. Please review.

Comparison is base (6d17662) 90.88% compared to head (06981fa) 85.91%.
Report is 13 commits behind head on develop.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #2337      +/-   ##
===========================================
- Coverage    90.88%   85.91%   -4.97%     
===========================================
  Files          492      492              
  Lines        45115    44674     -441     
===========================================
- Hits         41001    38380    -2621     
- Misses        4114     6294    +2180     

Files	Coverage Δ
nncf/common/graph/layer_attributes.py	96.26% <100.00%> (+0.03%)	⬆️
nncf/common/tensor_statistics/collectors.py	88.97% <100.00%> (-6.25%)	⬇️
nncf/openvino/statistics/collectors.py	95.32% <100.00%> (+0.71%)	⬆️
nncf/tensorflow/tensor_statistics/collectors.py	89.86% <100.00%> (ø)
nncf/torch/tensor_statistics/collectors.py	96.04% <95.83%> (+0.17%)	⬆️
nncf/onnx/statistics/collectors.py	0.00% <0.00%> (-80.92%)	⬇️

... and 49 files with indirect coverage changes

Flag	Coverage Δ
COMMON	44.83% <100.00%> (+0.36%)	⬆️
ONNX	?
OPENVINO	39.16% <32.05%> (-0.72%)	⬇️
TENSORFLOW	29.60% <5.12%> (+0.24%)	⬆️
TORCH	62.41% <32.05%> (+0.43%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Components	Coverage Δ
common	93.15% <100.00%> (-0.43%)	⬇️
torch	93.28% <95.83%> (ø)
tensorflow	94.01% <100.00%> (ø)
onnx	0.00% <0.00%> (-97.01%)	⬇️
openvino	90.64% <100.00%> (-1.24%)	⬇️
ptq	82.18% <ø> (-7.59%)	⬇️