Guard rcp rounded implementation as well

Since rcp implementations of non-default rounded versions are not correct or supported in OCML, guard them using the same macro OCML_BASIC_ROUNDED_OPERATIONS. Also update the docs and tests.


[ROCm/clr commit: 7b3bbc85c5]

projects/clr/hipamd/tests/src/deviceLib/hipDoublePrecisionIntrinsics.cpp

@@ -47,12 +47,10 @@ __device__ void double_precision_intrinsics() {
     __dmul_rn(1.0, 2.0);
     __dmul_ru(1.0, 2.0);
     __dmul_rz(1.0, 2.0);
-#endif
     __drcp_rd(2.0);
     __drcp_rn(2.0);
     __drcp_ru(2.0);
     __drcp_rz(2.0);
-#if defined OCML_BASIC_ROUNDED_OPERATIONS
     __dsqrt_rd(4.0);
     __dsqrt_rn(4.0);
     __dsqrt_ru(4.0);