Fix for scikit sparse dependency issue

[af-a]

Summary

Removing the scikit-sparse dependency.

TODOs

• Update README with note mentioning that scikit-sparse must be installed manually, if an optional flag for using Chloesky decomposition is specified (soon to be integrated).

Description

The scikit-sparse dependency was previously included in setup.py and requirements.txt.
However, installing LaPy using pip when scikit-sparse is specified as a dependency leads to errors:

Building wheels for collected packages: scikit-sparse
  Building wheel for scikit-sparse (PEP 517) ... error
  ERROR: Command errored out with exit status 1:
   command: /opt/conda/bin/python /opt/conda/lib/python3.8/site-packages/pip/_vendor/pep517/in_process/_in_process.py build_wheel /tmp/tmpfo7yw9fp
       cwd: /tmp/pip-install-w717c8s7/scikit-sparse_8636ab45e8894133b2998e2d9b87d24f
  Complete output (35 lines):
  running bdist_wheel
  running build
  running build_py
  creating build
  creating build/lib.linux-x86_64-3.8
  creating build/lib.linux-x86_64-3.8/sksparse
  copying sksparse/__init__.py -> build/lib.linux-x86_64-3.8/sksparse
  copying sksparse/test_cholmod.py -> build/lib.linux-x86_64-3.8/sksparse
  creating build/lib.linux-x86_64-3.8/sksparse/test_data
  copying sksparse/test_data/illc1033.mtx.gz -> build/lib.linux-x86_64-3.8/sksparse/test_data
  copying sksparse/test_data/illc1033_rhs1.mtx.gz -> build/lib.linux-x86_64-3.8/sksparse/test_data
  copying sksparse/test_data/illc1850.mtx.gz -> build/lib.linux-x86_64-3.8/sksparse/test_data
  copying sksparse/test_data/illc1850_rhs1.mtx.gz -> build/lib.linux-x86_64-3.8/sksparse/test_data
  copying sksparse/test_data/well1033.mtx.gz -> build/lib.linux-x86_64-3.8/sksparse/test_data
  copying sksparse/test_data/well1033_rhs1.mtx.gz -> build/lib.linux-x86_64-3.8/sksparse/test_data
  copying sksparse/test_data/well1850.mtx.gz -> build/lib.linux-x86_64-3.8/sksparse/test_data
  copying sksparse/test_data/well1850_rhs1.mtx.gz -> build/lib.linux-x86_64-3.8/sksparse/test_data
  running build_ext
  building 'sksparse.cholmod' extension
  creating build/temp.linux-x86_64-3.8
  creating build/temp.linux-x86_64-3.8/sksparse
  gcc -pthread -B /opt/conda/compiler_compat -Wl,--sysroot=/ -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -fPIC -Isksparse -I/tmp/pip-build-env-ozzw5wiz/overlay/lib/python3.8/site-packages/numpy/core/include -I/opt/conda/include -I/usr/include/suitesparse -I/opt/conda/include/python3.8 -c sksparse/cholmod.c -o build/temp.linux-x86_64-3.8/sksparse/cholmod.o
  In file included from /tmp/pip-build-env-ozzw5wiz/overlay/lib/python3.8/site-packages/numpy/core/include/numpy/ndarraytypes.h:1830:0,
                   from /tmp/pip-build-env-ozzw5wiz/overlay/lib/python3.8/site-packages/numpy/core/include/numpy/ndarrayobject.h:12,
                   from /tmp/pip-build-env-ozzw5wiz/overlay/lib/python3.8/site-packages/numpy/core/include/numpy/arrayobject.h:4,
                   from sksparse/cholmod.c:713:
  /tmp/pip-build-env-ozzw5wiz/overlay/lib/python3.8/site-packages/numpy/core/include/numpy/npy_1_7_deprecated_api.h:17:2: warning: #warning "Using deprecated NumPy API, disable it with " "#define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION" [-Wcpp]
   #warning "Using deprecated NumPy API, disable it with " \
    ^~~~~~~
  In file included from sksparse/cholmod.c:715:0:
  sksparse/cholmod_backward_compatible.h:1:10: fatal error: cholmod.h: No such file or directory
   #include "cholmod.h"
            ^~~~~~~~~~~
  compilation terminated.
  error: command '/usr/bin/gcc' failed with exit code 1
  ----------------------------------------
  ERROR: Failed building wheel for scikit-sparse
Failed to build scikit-sparse
ERROR: Could not build wheels for scikit-sparse which use PEP 517 and cannot be installed directly

Therefore, it was removed from the list of dependencies in both files.

In addition, the requirements.in file from which the requirements.txt file is generated (using the pip-compile tool) has also been included.

[af-a]

The following can be added as a note under the Installation section of the README:

The Solver can either use LU decomposition (default) or Cholesky decomposition if the use_cholmod flag is used. the latter depends on the scikit-sparse library. This library can not be installed among the depedencies of LaPy using pip (if it is specified in setup.py or requirements.txt), since it causes errors. Therefore, if you would like to use use_cholmod, you need to install scikit-sparse separately.

@m-reuter is this description sufficient?

[m-reuter]

Use this Note:

Several functions, e.g. the Solver, require a sparse matrix decomposition, for which either the LU decomposition (from scipy sparse) or the faster Cholesky decomposition (from scikit-sparse cholmod) can be used. If the parameter flag use_cholmod is True, the code will try to import cholmod from the scikit-sparse package and will fall back to LU if the import fails. If you would like to use cholmod, you need to install scikit-sparse separately. It cannot be listed among LaPy's dependencies (e.g. in setup.py or requirements.txt) as that causes errors with pip. scikit-sparse requires numpy and scipy to be installed separately beforehand.