Fix confusing width in simd_load (#4714)
I found the width using in simd_load is different from simd_store.
This implementation confuses me.
The reason lies in the missing parentheses for the type conversion of x
in the SIMD_LOAD2 macro definition, disrupting the intended semantics of
width variable.
I try to make a quick fix for it.
Co-authored-by: Logan Adams <114770087+loadams@users.noreply.github.com>
