fix: add unix_epoch_second() with floor semantics for negative timestamps

[0xSoftBoi]

Summary

• jiff::Timestamp::as_second() uses truncation toward zero, so @-1.5 gives -1 instead of the GNU-expected -2 (floor toward negative infinity)
• Added ParsedDateTime::unix_epoch_second() that corrects for negative subsecond components by subtracting 1 from the truncated second when subsec_nanosecond() < 0
• Added ParsedDateTime::subsec_nanosecond() that re-normalizes to the non-negative [0, 1_000_000_000) range matching the GNU timespec convention
• The uutils date command can use these methods to produce correct %s output

Fixes#283

Test plan

• 7 new tests covering positive, negative, fractional, zero, and large-value epoch timestamps
• All 382 existing tests pass (cargo test)
• Verified the exact examples from the issue: @-1.5 → -2, @-893375784.554767216 → -893375785

[codspeed-hq]

Merging this PR will not alter performance

✅ 21 untouched benchmarks

---

_{Comparing 0xSoftBoi:fix/negative-fractional-epoch-floor (27b7360) with main (4b412e0)}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.32%. Comparing base (4b412e0) to head (27b7360).

Additional details and impacted files

@@ Coverage Diff @@## main #285 +/- ##
==========================================
+ Coverage 99.30% 99.32% +0.01% 
==========================================
Files 20 20 Lines 3894 3982 +88 Branches 122 124 +2 ==========================================
+ Hits 3867 3955 +88 
Misses 26 26 Partials 1 1 

Flag	Coverage Δ
macos_latest	99.32% <100.00%> (+0.01%)	⬆️
ubuntu_latest	99.32% <100.00%> (+0.01%)	⬆️
windows_latest	13.73% <0.00%> (-0.32%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 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.

[sylvestre]

Seems that 2 lines aren't covered by the test

[tomagsx]

Review bump here as well. The main actionable feedback seems to be the two uncovered lines plus the CodSpeed warning. I’m happy to tighten that up if needed, but I’d also appreciate a maintainer pass on whether the GNU-compatible floor behavior is the right shape for the library before I keep iterating.

[0xSoftBoi]

Addressed the coverage feedback: the remaining uncovered line on the patch was the panic arm of the let-else in the extended accessor test. Swapped it for a matches! + assert! pair so the invariant is still checked but the patch lands at 100%. CI should now be green modulo the CodSpeed perf flag.