fix: parse offset seconds in FromStr so Display round-trips

[nvxbug]

FixedOffset's Display and Debug print a sub-minute offset as +HH:MM:SS, for example +00:00:30, but scan::timezone_offset only reads +HH:MM. That scanner backs FromStr for both FixedOffset and DateTime (through parse_rfc3339_relaxed), so a value chrono prints cannot be parsed back: DateTime::from_str rejects its own to_string() with TooLong on the trailing :SS, and FixedOffset::from_str silently drops the seconds and returns a truncated offset.

The fix teaches timezone_offset to consume an optional seconds component behind a new allow_seconds flag, applying the sign inside the scanner so negative offsets like -00:00:30 stay correct, and turns it on only for the two FromStr paths. Keeping it in the scanner fixes both round-trips in one place while strict parse_from_rfc3339, parse_from_rfc2822, and the %z family keep rejecting seconds. Regression tests cover the FixedOffset and DateTime paths across positive and negative sub-minute offsets.

[codecov]

Codecov Report

❌ Patch coverage is 93.10345% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.01%. Comparing base (3ffcd1b) to head (aef386c).

Files with missing lines	Patch %	Lines
src/format/scan.rs	86.66%	2 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1801   +/-   ##
=======================================
  Coverage   90.01%   90.01%           
=======================================
  Files          35       35           
  Lines       13831    13855   +24     
=======================================
+ Hits        12450    12472   +22     
- Misses       1381     1383    +2     

☔ 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.

[djc]

I'm going to stop responding to these PRs for edge case bugs. There are probably a lot more; you should direct your efforts towards other implementations.

[nvxbug]

Makes sense, I'll redirect to other things. Happy to close this if you'd rather not keep it on the pile, otherwise it's here if the round-trip ever matters.