Improve Tensor performance

[AlexanderDokuchaev]

Changes

Add custom custom implementation of functools.singledispatch with custom arguments 'wrapper_type' to select wrapper function. Remove np.errstate from divide function because it degrades performance, to disable warning add disable_error_handling function.

Tested on run 1000 times tune_range function by cProfile.

	calls	time
develop	338187	0.167
remove np.errstate from div	262113	0.127
tensor_dispatch	241006	0.118

[codecov[bot]]

Codecov Report

Merging #2255 (76654ca) into develop (51aac27) will decrease coverage by 0.04%. Report is 1 commits behind head on develop. The diff coverage is 88.52%.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #2255      +/-   ##
===========================================
- Coverage    90.70%   90.67%   -0.04%     
===========================================
  Files          485      486       +1     
  Lines        43601    43600       -1     
===========================================
- Hits         39549    39535      -14     
- Misses        4052     4065      +13     

Flag	Coverage Δ
COMMON	15.72% <84.42%> (-0.03%)	:arrow_down:
ONNX	33.80% <79.50%> (-0.02%)	:arrow_down:
OPENVINO	38.57% <79.50%> (-0.02%)	:arrow_down:
TENSORFLOW	30.05% <63.93%> (+<0.01%)	:arrow_up:
TORCH	62.61% <88.52%> (-0.04%)	:arrow_down:

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

Files	Coverage Δ
nncf/experimental/tensor/functions.py	100.00% <100.00%> (+0.75%)	:arrow_up:
nncf/experimental/tensor/numpy_functions.py	100.00% <100.00%> (ø)
nncf/experimental/tensor/tensor.py	99.00% <100.00%> (-0.01%)	:arrow_down:
nncf/experimental/tensor/torch_functions.py	98.01% <100.00%> (-0.18%)	:arrow_down:
nncf/quantization/fake_quantize.py	94.78% <100.00%> (+0.04%)	:arrow_up:
nncf/experimental/tensor/dispatcher.py	82.27% <82.27%> (ø)

... and 1 file with indirect coverage changes