Fix timestamp96 decoding at zero seconds

[fhgffy]

A timestamp96 value with seconds=0 and nanoseconds=500 currently decodes to 1 microsecond instead of 0 through both as and convert. With a seconds-resolution destination, it becomes 1 second. The equivalent timestamp64 representation correctly truncates to 0.

The timestamp96 decoder sends sec == 0 through the negative-second adjustment. Change both conditions to sec >= 0, so nonnegative fractions use the same truncation direction as timestamp64. Negative-second behavior is unchanged.

Add a regression with explicit timestamp96 bytes and a microsecond-resolution time_point, covering both public conversion paths. Using an explicit duration avoids hiding the bug on hosts whose system clock has nanosecond precision.

Fresh validation on Linux x86-64 with GCC 14.2.0 and Boost 1.85.0:

• Added regression fails on the unmodified base with two assertions and passes with the fix
• Full C++11 CTest suite: 39/39 passed
• Targeted C++20 msgpack_cpp11 and object_with_zone tests with UBSan: 2/2 passed
• Separate edge checks pass under C++11 and C++20, API versions 1/2/3, warning-as-error flags, and UBSan; equivalent timestamp32/64/96 values and negative/zero/positive neighboring seconds are covered
• Independent review found no blocking issues; a separate deterministic property test passes 841,152 assertions per configuration across six duration types

Focused test: ./build/test/msgpack_cpp11 --run_test=system_clock_96_zero_seconds

Full C++20, non-Linux platforms, and out-of-range destination durations were not tested. Timestamp96 is a legal representation for these values under the timestamp specification, even when a serializer could choose a shorter format.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 79.91%. Comparing base (39cc081) to head (0f0fdfb).
⚠️ Report is 24 commits behind head on cpp_master.
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@              Coverage Diff               @@
##           cpp_master    #1190      +/-   ##
==============================================
- Coverage       85.36%   79.91%   -5.46%     
==============================================
  Files              80       80              
  Lines            5099     5790     +691     
  Branches          507      629     +122     
==============================================
+ Hits             4353     4627     +274     
- Misses            746     1163     +417     

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[redboltz]

@fhgffy , thank you for sending the PR. It looks good to me.