v: some branch prediction adjusts

[felipensp]

This PR aims improve performance by doing branch prediction adjusts.

Reduced 26.004.738 (Bcm) to 25.334.846 (Bcm) on v -o cmd.c cmd/v

Bcm = branch predictor misses

$ v -cflags -O0 -cc gcc -cg -cflags -g self
$ valgrind --tool=callgrind --branch-sim=yes --cache-sim=yes  --dump-instr=yes ./v -no-parallel -o cmd.c cmd/v
$ callgrind_annotate --show=Bc,Bcm --sort=Bcm --tree=both --inclusive=no --auto=yes --context=1 callgrind.out.17213 > tree.17213

_{Huly®: V_0.6-21293}

[spytheman]

Excellent work.

[spytheman]

Note that compiling with -O0 -cc gcc will turn off all optimizations, including the ones that turn switches into jump tables.

I am not convinced that doing that, and then reordering code around, just to improve that particular mode is good 🤔 , except perhaps for tcc builds (which also does not do optimizations).