Address short comings with printf for strings

[Rekt3421]

This change fixes short comings encountered in 9 tests of the OpenCL-CTS as mentioned in issue #714 we still need to fix kernel descriptors.

[Rekt3421]

Next step is to add some unit tests

[rjodinchr]

This is fixing CTS printf string, but printf format_string is regressing (segfault with "\"%%\"").

We should maybe try to address everything (string & format_string) together.

[Rekt3421]

current status 3 sub-subtests are still failing. One which is due to some issue with printf itself.

I.e. in strings subtest there is a subtest

printf(%%foo%%) 

is printing out

%%foo%

which is a bit confusing since the two %% should be consumed as one.

also in the same subtests for

   printf("%s\n","foo\0foo");                                                           

the kernel descriptors are not properly populated for some reason.

a vector sub-subtest is also failing for which I do not have an explanation yet.

[Rekt3421]

@kpet to get this PR working correctly we need to update clspv with the fixes landed by David