fix calculate_relative_offset year offset from Feb 29

[ThomasWaldmann]

Fixes #9967.

Routes the y case in calculate_relative_offset through offset_n_months(from_ts, offset * 12) as suggested in the issue. Result: Feb 29 + Ny no longer crashes when the target year is not a leap year — the day is clamped to the target month's last day (Feb 28 in a non-leap year, Feb 29 in a leap year).

Also extended offset_n_months to preserve hour/minute/second/microsecond of from_ts — otherwise routing the year case through it would have silently dropped time-of-day (and the month case had the same latent gap).

Regression tests added in src/borg/testsuite/helpers/time_test.py.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 85.72%. Comparing base (a50eb6f) to head (471f451).
⚠️ Report is 8 commits behind head on master.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #9968      +/-   ##
==========================================
+ Coverage   85.70%   85.72%   +0.02%     
==========================================
  Files          95       95              
  Lines       16815    16845      +30     
  Branches     2577     2581       +4     
==========================================
+ Hits        14411    14441      +30     
  Misses       1668     1668              
  Partials      736      736              

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

[PhrozenByte]

Also extended offset_n_months to preserve hour/minute/second/microsecond of from_ts

In case a borg2 beta tester notices a regression: this slightly changes the behavior of borg2 repo-list --oldest/--newest/--older/--newer (these selectors don't exist in Borg 1). Previously, running borg2 repo-list --newer 1m (assuming "now" is 2026-07-29 22:00:00) would also include an archive created at 2026-06-29 00:00:00. This is no longer the case, the cut-off date is now 2026-06-29 22:00:00 instead of 2026-06-29 00:00:00.

Since borg2 is still in beta and the new behavior better matches the behavior of all other intervals, this is just a heads-up. IMHO there's no need to change or document this.