Fix issues in handle_func_command in syft tensors

[LaRiffle]

Description

This cleans code in handle_func_command and fix bugs due to variable overwriting for syft tensors, and to overloading functions in native.py

[gmuraru]

Q: Could you add a test for this?

[codecov]

Codecov Report

Merging #3807 into master will increase coverage by 0.00%.
The diff coverage is 97.05%.

@@           Coverage Diff            @@
##           master    #3807    +/-   ##
========================================
  Coverage   94.69%   94.70%            
========================================
  Files         187      187            
  Lines       18626    18961   +335     
========================================
+ Hits        17638    17957   +319     
- Misses        988     1004    +16     

Impacted Files	Coverage Δ
...ft/frameworks/torch/tensors/interpreters/native.py	91.25% <92.30%> (-0.03%)	⬇️
...orks/torch/tensors/interpreters/additive_shared.py	93.43% <100.00%> (+0.01%)	⬆️
.../frameworks/torch/tensors/interpreters/autograd.py	88.23% <100.00%> (+0.12%)	⬆️
...frameworks/torch/tensors/interpreters/precision.py	97.26% <100.00%> (+0.01%)	⬆️
syft/execution/translation/threepio.py	91.04% <0.00%> (-8.96%)	⬇️
syft/execution/computation.py	96.15% <0.00%> (-3.85%)	⬇️
syft/frameworks/torch/he/fv/util/operations.py	90.97% <0.00%> (-2.58%)	⬇️
syft/generic/frameworks/attributes.py	87.75% <0.00%> (-1.61%)	⬇️
syft/exceptions.py	77.63% <0.00%> (-0.55%)	⬇️
syft/frameworks/torch/he/fv/evaluator.py	93.70% <0.00%> (-0.35%)	⬇️
... and 13 more

[LaRiffle]

Q: Could you add a test for this?

I've updated the code once more maybe it's clearer now. It's hard to add tests for this, but it's needed for batchnorm that I wanted to put in a separate PR and which is tested. Sounds good?
(and the main function at stakes here, torch.roll, is tested)

[gmuraru]

PICKY Stuff: Here the comment should be removed, right? Since we look for native_{cmd}

[LaRiffle]

haha no it's good to be picky. here the module native_torch defined in native.py overwrites native torch functions... Yeah it's a bit strange, I really had to modify a torch function (aka torch.roll) to behave slightly differently

[gmuraru]

LGTM! Only one small question