Fix Float::copysign impl

[Robbepop]

I discovered today that the official WebAssembly spec test suite assertions and wasmi results diverge in a particular case with the copysign instruction.
For F32::from_bits(0xFFFC_0000).copysign(F32::from_bits(0)) the expected result according to the WebAssembly Spec test suite is 0x7FFC_0000, however, wasmi currently returns 0xFFFC_0000.
The Rust documentation mandates the same behavior from copysign as WebAssembly does: https://doc.rust-lang.org/std/primitive.f32.html#method.copysign

The bug was not caught because the current wasmi Wasm spec test runner seems to improperly handle NaN values.

Fortunately the bug fix was easy, however, I still wonder why we are going to great lengths like this.

[codecov-commenter]

Codecov Report

Merging #293 (ac18ee5) into master (8b27c2b) will increase coverage by 0.01%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #293      +/-   ##
==========================================
+ Coverage   80.00%   80.02%   +0.01%     
==========================================
  Files          30       30              
  Lines        4672     4676       +4     
==========================================
+ Hits         3738     3742       +4     
  Misses        934      934              

Impacted Files	Coverage Δ
src/value.rs	69.45% <100.00%> (+0.61%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8b27c2b...ac18ee5. Read the comment docs.

[athei]

The bug was not caught because the current wasmi Wasm spec test runner seems to improperly handle NaN values.

Shouldn't we fix this first then? To make sure that this does not brake other things.

[Robbepop]

Shouldn't we fix this first then? To make sure that this does not brake other things.

The v1 implementation caught this bug since v1 properly tests this property.
The v0 implementation of the Wasm spec test builds on wabt whereas v1 builds on wast crate.
The fix is very isolated - we are just touching the one particular float operation.
I agree that ideally we should also fix the Wasm spec test suite for v0 but this should best be done in another PR imo.

[athei]

This code is used by v0 and v1?

[Robbepop]

This code is used by v0 and v1?

yes, v1 already implements the fix. This PR backports it to v0.
Same for #295 in case you wonder.