explicitly define dtype for sparse matrix

[klausweinbauer]

Fix issue #113 by setting the datatype to float during sparse matrix initialization.

[ysig]

Thank you @klausweinbauer Is this backwards compatible? If not please add a flag.