Problem setting matrixtype in pyculib.sparse.Sparse.matdescr

I have a problem with the matrix descriptor class, when giving parameters different from default ones :

``` python
>>> import pyculib
>>> handle = pyculib.sparse.Sparse()
>>> handle

>>> descr = handle.matdescr()
>>> descr.matrixtype
'G'
>>> descrS = handle.matdescr(matrixtype='S')
>>> descrS.matrixtype
Traceback (most recent call last):
File "", line 1, in
File "~/anaconda3/envs/OTML/lib/python3.6/site-packages/pyculib/sparse/binding.py", line 201, in matrixtype
return MATRIXTYPECHAR[self._api.cusparseGetMatType(self._handle)]
File "~/anaconda3/envs/OTML/lib/python3.6/site-packages/pyculib/utils/libutils.py", line 48, in wrapped
self.check_error(status)
File "~/anaconda3/envs/OTML/lib/python3.6/site-packages/pyculib/utils/libutils.py", line 54, in check_error
raise self.ErrorType(status)
pyculib.sparse.binding.CuSparseError: CUSPARSE_STATUS_NOT_INITIALIZED

>>> descrT = handle.matdescr(matrixtype='T')
>>> descrT.matrixtype
Traceback (most recent call last):
File "", line 1, in
File "/home/matthieu/anaconda3/envs/OTML/lib/python3.6/site-packages/pyculib/sparse/binding.py", line 201, in matrixtype
return MATRIXTYPECHAR[self._api.cusparseGetMatType(self._handle)]
File "/home/matthieu/anaconda3/envs/OTML/lib/python3.6/site-packages/pyculib/utils/libutils.py", line 48, in wrapped
self.check_error(status)
File "/home/matthieu/anaconda3/envs/OTML/lib/python3.6/site-packages/pyculib/utils/libutils.py", line 54, in check_error
raise self.ErrorType(status)
pyculib.sparse.binding.CuSparseError: CUSPARSE_STATUS_INVALID_VALUE

>>> descrH = handle.matdescr(matrixtype='H')
>>> descrH.matrixtype
Traceback (most recent call last):
File "", line 1, in
File "/home/matthieu/anaconda3/envs/OTML/lib/python3.6/site-packages/pyculib/sparse/binding.py", line 201, in matrixtype
return MATRIXTYPECHAR[self._api.cusparseGetMatType(self._handle)]
File "/home/matthieu/anaconda3/envs/OTML/lib/python3.6/site-packages/pyculib/utils/libutils.py", line 48, in wrapped
self.check_error(status)
File "/home/matthieu/anaconda3/envs/OTML/lib/python3.6/site-packages/pyculib/utils/libutils.py", line 54, in check_error
raise self.ErrorType(status)
pyculib.sparse.binding.CuSparseError: CUSPARSE_STATUS_ALLOC_FAILED

```

Any ideas of what could be happening ? I tried to debug it, but the function that returns the error is `self._api.cusparseGetMatType` and I can't access inside.

Other than that, I am also confused by the [matrix descriptor](https://docs.nvidia.com/cuda/cusparse/index.html#cusparsematrixtypet) documentation in CUDA. The text says:

> This type indicates the type of matrix stored in sparse storage. Notice that for symmetric, Hermitian and triangular matrices only their lower or upper part is assumed to be stored.
>
> The whole idea of matrix type and fill mode is to keep minimum storage for symmetric/Hermitian matrix, and also to take advantage of symmetric property on SpMV (Sparse Matrix Vector multiplication). To compute y=A\*x when A is symmetric and only lower triangular part is stored, two steps are needed. First step is to compute y=(L+D)\*x and second step is to compute y=L^T\*x + y. Given the fact that the transpose operation y=L^T\*x is 10x slower than non-transpose version y=L\*x, the symmetric property does not show up any performance gain. It is better for the user to extend the symmetric matrix to a general matrix and apply y=A\*x with matrix type CUSPARSE_MATRIX_TYPE_GENERAL.

Does that mean I should just use the General matrix type even if my matrix is symmetric ? But then why say that `The whole idea of matrix type and fill mode is to keep minimum storage for symmetric/Hermitian matrix, **and also to take advantage of symmetric property on SpMV**` ?

I know that this might be more a CUDA question than a pyculib, but maybe someone has the answer...

My config:

CUDA 8.0
Ubuntu 16.04
pyculib 1.0.2+2.g7ae9662
```
$ conda info
active environment : None
...
conda version : 4.5.12
conda-build version : 3.0.27
python version : 3.6.8.final.0
...
platform : linux-64
user-agent : conda/4.5.12 requests/2.18.4 CPython/3.6.8 Linux/4.4.0-141-generic ubuntu/16.04 glibc/2.23
UID:GID : 1000:1000
netrc file : None
offline mode : False
```

Start with pyculib/sparse/binding.py, especially matdescr and the matrixtype property, then inspect error handling in pyculib/utils/libutils.py. Reproduce the descriptor calls with the reported CUDA 8.0 configuration and compare each requested matrix type with the CUDA API's accepted values; done means valid types behave consistently and invalid requests return an understandable result.

Written by the indexing model from the issue text.