[BUG] SIMD Reduce functions don't work at compile time

[martinvuyk]

Bug description

Actual behavior

fn main():
    alias a = SIMD[DType.uint8, 2](0, 1).reduce_and()
    alias b = SIMD[DType.uint8, 2](0, 1).reduce_or()
    alias c = SIMD[DType.uint8, 2](0, 1).reduce_min()
    alias d = SIMD[DType.uint8, 2](0, 1).reduce_max()
    alias e = SIMD[DType.uint8, 2](0, 1).reduce_max()
    print(a, b, c, d, e)

output:

note: failed to interpret operation pop.call_llvm_intrinsic{fastmathFlags: #pop<fmf none>, hasSideEffects: false, intrin: "llvm.vector.reduce.and" : !kgen.string}(#kgen.pack<<0, 1>> : !kgen.pack<[[typevalue<#kgen.typeref<@"stdlib::builtin::simd::SIMD,dtype=ui8,size=2">>, simd<2, ui8>]]>)

Steps to reproduce

...

System information

Version mojo 25.3.0.dev2025032905

---

Part of #4142

CC: @lattner