Mixed precision: Provide conversion routines

[karlrupp]

Currently one cannot create a vector of integers from a vector of floats or vice versa. However, this would allow for some neat mixed precision algorithms, so we should provide conversions of the type

viennacl::vector x(viennacl::element_cast(x_int));

or

viennacl::vector<float> x(x_int);

to better support this. See also discussion here:
http://sourceforge.net/p/viennacl/mailman/viennacl-devel/thread/CAFXwtss7-%2BA_4w7bNw3jbdZ1xn8%2BAxo_71K8igO%2Bg7qqyRGJCA%40mail.gmail.com/#msg32493195

[karlrupp]

Postponed to 1.7.0 in order to install shorter release cycles.