Interpreter: Don't change NaN bits when dividing by 1

[kripken]

It's valid to change NaN bits in that case per the wasm spec, but
if we do so then fuzz testcases will fail on the optimization of

nan:foo / 1 => nan:foo

That is, it is ok to leave the bits as they are, and if we do that then
we are consistent with the simple and valid optimization of removing
a divide by 1.

Found by the fuzzer - looks like on x64 on some float32 NaNs,
the bits will actually change (see the testcase). I've seen this on
two machines consistently, so it's normal apparently.

[kripken]

Huh, this fails a spec test,

;; Test that x/1.0 is not folded to x.

(module
  (func (export "f32.no_fold_div_one") (param $x f32) (result f32)
    (f32.div (local.get $x) (f32.const 1.0)))
  (func (export "f64.no_fold_div_one") (param $x f64) (result f64)
    (f64.div (local.get $x) (f64.const 1.0)))
)

(assert_return (invoke "f32.no_fold_div_one" (f32.const nan:0x200000)) (f32.const nan:0x600000))
(assert_return (invoke "f64.no_fold_div_one" (f64.const nan:0x4000000000000)) (f64.const nan:0xc000000000000))

Does anyone understand what that test is doing? I'm puzzled by the nan bits change. It seems I'm missing something important here.

[kripken]

Ah, the spec test has been fixed in upstream, to

(assert_return (invoke "f32.no_fold_div_one" (f32.const nan:0x200000)) (f32.const nan:arithmetic))
(assert_return (invoke "f64.no_fold_div_one" (f64.const nan:0x4000000000000)) (f64.const nan:arithmetic))

We don't support that new test notation here, but anyhow I'll just remove that particular test, as the new test added here is even stricter.