Pyitlib elimination

[Roman223]

Package conflict
Pyitlib requires scikit-learn lower than 1.0.0.
The problem is that old versions have troubles with Cython on some machines and we decided to move pyitlib into external and import it as a build-in package bypassing pip installation. It turns out that part we need contains the only one object from sklearn - LabelEncoder, which is stable across mostly versions.

Tests imrovements
Sometimes pickle compression works and a test returns error. This case was covered.

Code formatted to black style as well.

[codecov-commenter]

Codecov Report

Merging #70 (2580ede) into master (7e11513) will increase coverage by 0.04%.
The diff coverage is 33.33%.

@@            Coverage Diff             @@
##           master      #70      +/-   ##
==========================================
+ Coverage   65.11%   65.15%   +0.04%     
==========================================
  Files          32       32              
  Lines        2250     2250              
==========================================
+ Hits         1465     1466       +1     
+ Misses        785      784       -1     

Files Changed	Coverage Δ
bamt/networks/base.py	48.62% <33.33%> (ø)

... and 5 files with indirect coverage changes