fix(math): fix double rounding in Math.f16round

[ParthMozarkar]

Hey! I noticed Math.f16round was failing the test262 value-conversion.js
test and dug into why.

The issue is that the float16 crate's from_f64 converts through f32
internally, which causes double rounding for certain values.

For example Math.f16round(1.00048828125000022204) was returning 1
instead of the correct 1.0009765625.

I fixed it by switching to the half crate's f16::from_f64 which is
already in the dependency tree and converts directly from f64 to f16
without any intermediate f32 conversion.

All existing tests pass.

Fixes #5237

[github-actions]

Test262 conformance changes

Test result	main count	PR count	difference
Total	52,963	52,963	0
Passed	50,545	50,545	0
Ignored	1,426	1,426	0
Failed	992	992	0
Panics	2	2	0
Conformance	95.43%	95.43%	0.00%

Tested main commit: 66a1cd268ce6006918032b5d0f98d3a63c81b188
Tested PR commit: a76e227ce36c8ccc2ade29523eacad1e5f160c78
Compare commits: 66a1cd2...a76e227

[codecov]

Codecov Report

❌ Patch coverage is 0% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 59.88%. Comparing base (6ddc2b4) to head (a76e227).
⚠️ Report is 942 commits behind head on main.

Files with missing lines	Patch %	Lines
core/engine/src/builtins/math/mod.rs	0.00%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #5238       +/-   ##
===========================================
+ Coverage   47.24%   59.88%   +12.64%     
===========================================
  Files         476      582      +106     
  Lines       46892    63460    +16568     
===========================================
+ Hits        22154    38006    +15852     
- Misses      24738    25454      +716     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[zhuzhu81998]

switching to half does not solve the problem.
on ur branch:

because half seems to do the same thing. the reason is probably what the float16 comment said in the FIXME:

#[inline]
#[target_feature(enable = "f16c")]
pub(super) unsafe fn f64_to_f16_x86_f16c(f: f64) -> u16 {
    // FIXME: Change to use `_mm_cvtpd_ph` when stable.
    f32_to_f16_x86_f16c(f as f32)
}

[hansl]

Correct. float16 is a fork of half because half had too many dependencies (crunchy and zerocopy). There is also a PR on float16 that might fix this: Alexhuszagh/float16#16

Have you tried it?

[jedel1043]

Closing since this doesn't fix the issue, and the issue seems to be fixed by a PR upstream.