[TargetLowering][RISCV] Improve codegen for saturating bf16 to int conversion

Extending to f32 first (as is done for f16) results in better generated
code for RISC-V (and affects no other in-tree tests). Additionally,
performing the FP_EXTEND first seems equally justified for bf16 as for
f16.

Differential Revision: https://reviews.llvm.org/D156944