New C++ (Intrinsics only) version.
Allows for mixed precision dslashery. Eg:

Dslash<float>
Dslash<double>
Dslash3D<float>
Dslash3D<double>

operator() does all the applications.

Gauge field still needs to be packed... (Work on this?)
Otherwise interface is the same/similar to old SSE Dslash

To Do:
 - straight C versions of the SSE for testing on Intel/Pathscale compiler
   also for loop unrollings etc etc etc. This primarily needs rewrite
   of the decompose reconstruct (TEDIOUS but useful?)

 - autoconf setting of details in cache.h 

 - bells and whistles
	* funky data orderings. 
	* combine with clover for high level clover op?
 	* temp prec C_L and C_R backsubstitutions (should be easy?)