BUG: make resample with Day freq consistent with equivalent Hour freq

[jbrockmendel]

Summary

• Fixes an inconsistency where resample("7D") produced different bin edges than resample("168h"), even though 7 days = 168 hours (GH#44996)
• Day freq now goes through _adjust_dates_anchored (same path as Tick offsets), with tz-stripping to handle DST correctly
• origin and offset parameters are now respected for Day frequencies instead of being silently ignored

Test plan

• Existing resample tests pass (4106 passed)
• New parametrized tests verify 7D matches 168h across closed/label combinations
• New parametrized test verifies origin works with Day freq across tz={None, UTC, US/Eastern}
• DST edge cases tested (nonexistent times, fall-back transitions)

Closes #44996

🤖 Generated with Claude Code