Moved getNonZeroRowLengthCudaKernal to SparseRow_impl.h and indentified errors in SparseRow_impl.h . Commiting for backup purposes.

Deleted useless code, reformatted present code. Still have issues with getNonZeroRowLength. getRow() throws SegFault and so does resultCuda. Commiting for backup purposes.

a row. Commiting for backup purposes.

Changed implementation of getting Non-Zero elements. Added problem to comments. Commiting for backup purposes.

Reformatted getNonZeroRowLength to pass on the device Type in string and thus enable the lambda in SparseRow_impl.h

Tried using std::vector to avoid having to implement DeviceType into SparseRow. Commiting for backup purposes.

Fixed mistake where integer division was used for certain format declarations. Added commented-out helpful prints to identify errors.

Fixed POSSIBLE MISTAKE in both addElement functions, where columns where being compared to this->rows, changed the formed to be compared to this->columns.

Fixed mistake with setSlice, ceil function used RealType and division, this would cause integer division at certain format declarations, now uses float only.

The former has been formally deprecated for a long time and in CUDA 10.0
it is finally properly marked as deprecated, so the compiler can print
warnings.

The x and b vectors might have different type.