Fix %I/%-I strftime directives to use the 12-hour clock

[gaoflow]

Problem

The %I and %-I strftime directives map straight to the hour attribute
(0–23), so they render identically to %H instead of a 12-hour clock:

>>> import jdatetime
>>> jdatetime.datetime(1400, 1, 1, 13, 0).strftime('%I')   # 1pm
'13'      # should be '01'
>>> jdatetime.datetime(1400, 1, 1, 0, 0).strftime('%I')    # midnight
'00'      # should be '12'

The '12' fallback already present in the mapping shows these are meant to be
the 12-hour clock. The reference is the stdlib: datetime.strftime('%I').

Fix

Add an hour12 property (hour % 12, with 0 mapped to 12) and point %I
and %-I at it. A plain date (no time component) still resolves to the '12'
fallback, so existing date behavior is unchanged.

Tests

test_strftime_I_directive_uses_twelve_hour_clock sweeps all 24 hours and
asserts %I / %-I match the stdlib output, checks a %p+%I round-trip
invariant (reconstructing the 24-hour value for every hour), spot-checks the
values the old behavior got wrong (0→'12', 13→'01', 23→'11'), and
the date-midnight→'12' fallback. Full suite: 127 passed, 3 skipped, 76
subtests. ruff check / ruff format --check clean.

[5j9]

Since jdatetime aims for parity with Python's standard datetime module, it might be a good idea to make the hour12 property private (e.g., _hour12).

[slashmili]

Thanks @5j9! I agree

[gaoflow]

Done — hour12 is now _hour12 (a37985b), and the branch is rebased onto main with the changelog entry moved to a new Unreleased section. All tests pass.