[Clang][RISCV] Add packed byte horizontal multiply-accumulate intrinsics (#234484)
Add `__riscv_pm4adda_i8x8`, `__riscv_pm4addau_u8x8`, and
`__riscv_pm4addasu_i8x8` from the Packed Multiplication with Horizontal
Addition and Accumulate section of the P-extension intrinsic
specification.
The wrappers reuse the existing packed dot-product intrinsics and
generic vector addition. RV64 selects one byte multiply-accumulate
instruction; RV32 selects one for each word.