Fixes for SPR/SPR2 BLAS 2 routines

[pgorlani]

This patch fixes various aspects of the SPR/SPR2 BLAS2 routines:

• fix the mapping of the global work-item index to the triangular matrix coordinates by using only integer arithmetic;
• move method definitions to the proper file;
• remove unnecessary increment arguments;
• improve unit tests by randomly generating the matrix and removing the unnecessary memory layout argument;
• improve documentation.

[muhammad-tanvir-1211]

LGTM, apart from the comment regarding layout in tests. Thanks.

[OuadiElfarouki]

Looks good thank you!