Updates for pytorch 1.6.0

[mtar]

Description

Update the unittests for argmax and argmin for the changed return behaviour (numpy like) resulting from changes in pytorch 1.6.0 (torch.max, torch.min return first value).
Additionally, mpi_argmax and mpi_argmin sort the inputs by starting index beforehand. Some MPI implementations do not bind the first input to the lower process.

Changes proposed:

• argmax test update
• argmin test update
• requires torch 1.6.0
• sort arguments in mpi_argmax, mpi_argmin (Fix bug in some MPI implementations)
• updated gpu calls in a few tests (see code for details)
• io tests use the current working directory for the tests instead of a temp directory

Type of change

• Bug fix (non-breaking change which fixes an issue)

Due Diligence

• All split configurations tested
• Multiple dtypes tested in relevant functions
• Documentation updated (if needed)
• Updated changelog.md under the title "Pending Additions"

Does this change modify the behaviour of other functions? If so, which?

Unknown

[coquelin77]

these tests will not run on GPU. is there a reason that we cant use the torch argmax/argmin for the tests?

[codecov]

Codecov Report

Merging #653 into master will decrease coverage by 0.02%.
The diff coverage is 95.00%.

@@            Coverage Diff             @@
##           master     #653      +/-   ##
==========================================
- Coverage   97.40%   97.37%   -0.03%     
==========================================
  Files          85       85              
  Lines       16636    16646      +10     
==========================================
+ Hits        16204    16209       +5     
- Misses        432      437       +5     

Impacted Files	Coverage Δ
heat/cluster/tests/test_kmedoids.py	100.00% <ø> (ø)
heat/core/printing.py	98.41% <75.00%> (-1.59%)	⬇️
heat/core/statistics.py	96.76% <77.77%> (-0.56%)	⬇️
heat/cluster/tests/test_spectral.py	100.00% <100.00%> (ø)
heat/core/_operations.py	92.73% <100.00%> (ø)
heat/core/dndarray.py	97.22% <100.00%> (-0.01%)	⬇️
heat/core/linalg/basics.py	94.46% <100.00%> (ø)
heat/core/manipulations.py	99.12% <100.00%> (ø)
heat/core/tests/test_dndarray.py	100.00% <100.00%> (ø)
heat/core/tests/test_io.py	93.03% <100.00%> (ø)
... and 3 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 7616ca8...cfb588b. Read the comment docs.

[coquelin77]

build failing -> due to issues with jenkins
codecov -> most changes are within aspects which are not touched (5 lines) by travis, 4 in MPI operations, 1 is a print change for old torch versions

[Cdebus]

Looks good to me and tests are passing