cmd/compile: BCE of fixed width blocks during iteration

[Jorropo]

With SIMD it is normal to iterate a slice let's say 8 elements at a time.
Sadly this incurs a bound check:

	const step = 8
	for i := 0; i <= len(x)-step; i += step {
		_ = x[i : i+step] // Unnecessary bound check here :'(
	}

[TapirLiu]

also for

	for i := 0; i <= len(x)-step; i += step {
		_ = x[i+step-1] // Unnecessary bound check here :'(
	}

see #40987