As we only store 32bit floats in daq data stream, only deal with floats.
Then remove double->float cast in the fast copy loop.
Update the fast copy loop to be a memcpy.



git-svn-id: https://redoubt.ligo-wa.caltech.edu/svn/advLigoRTS/branches/branch-3.5@4943 6dcd42c9-f523-4c6d-aada-af552506706e