add direct testing of CUDA block algorithms

[jaredhoberock]

The current block algorithms are listed below and organized into categories based on inputs and outputs.

X->X (inplace)
inclusive_scan
inclusive_scan_n
inclusive_scan_by_flag
inclusive_scan_by_flag_n
reduce

X,Y->X,Y (inplace)
merging_sort

X->Y
copy

X,Y->Z
merge

X,Y,Z->W
set_difference
set_intersection
set_symmetric_difference
set_union

Forwarded from http://code.google.com/p/thrust/issues/detail?id=406

[alliepiper]

No longer relevant.