RISC-V: Remove HF modes of FP to INT rounding autovec

RISC-V: Remove HF modes of FP to INT rounding autovec

The [i|l|ll][rint|round|ceil|floor] internal functions are
defined as DEF_INTERNAL_FLT_FN instead of DEF_INTERNAL_FLT_FLOATN_FN.
Then the *f16 (N=16 of FLOATN) format of these functions are not
available when try to get the ifn from the given cfn in the
vectorizable_call. Aka:

BUILT_IN_LRINTF16 => IFN_LAST (should be IFN_LRINT here)
BUILT_IN_RINTF16 => IFN_RINT

It is better to remove FP16 related modes until the additional
middle-end support is ready. This patch would like to clean the FP16
modes with some comments.

gcc/ChangeLog:

	* config/riscv/vector-iterators.md: Remove HF modes.

Signed-off-by: Pan Li <pan2.li@intel.com>