Fix rrule byweekno for days in the previous year's last ISO week

[vineethsaivs]

Summary

rrule(..., byweekno=...) misattributes the first days of a year that actually belong to the last ISO week of the previous year.

_iterinfo.rebuild() computes lnumweeks, the last week number of the previous year, to decide whether early-January spillover days should be marked for a given byweekno. It computed that value from the current year's length with a hardcoded 52:

if lno1wkst >= 4:
    lno1wkst = 0
    lnumweeks = 52+(lyearlen + (lyearweekday-rr._wkst) % 7) % 7//4
else:
    lnumweeks = 52+(self.yearlen-no1wkst) % 7//4   # uses THIS year's yearlen/no1wkst

The parallel current-year block a few lines above already does it correctly, using divmod on the week-adjusted length. Because the previous-year block used the wrong year's variables (and hardcoded 52 instead of the real div), lnumweeks came out one too high in some years.

Reproduction

2011-01-01 and 2011-01-02 are ISO week 52 of 2010 (2010 has no week 53), confirmed by date.isocalendar():

>>> from dateutil.rrule import rrule, YEARLY, SA, SU
>>> from datetime import datetime
>>> list(rrule(YEARLY, count=2, byweekno=52, byweekday=(SA, SU),
...            dtstart=datetime(2011, 1, 1, 9, 0)))
# before: [datetime(2011, 12, 31, 9, 0), datetime(2012, 1, 1, 9, 0)]   # skips Jan 1/2
# after:  [datetime(2011, 1, 1, 9, 0),  datetime(2011, 1, 2, 9, 0)]    # correct
>>> list(rrule(YEARLY, byweekno=53, dtstart=datetime(2011, 1, 1, 9, 0),
...            until=datetime(2011, 1, 10, 9, 0)))
# before: [datetime(2011, 1, 1, 9, 0), datetime(2011, 1, 2, 9, 0)]     # week 53 does not exist
# after:  []                                                            # correct

Fix

Compute the previous year's week count the same way the current-year block does, from the previous year's lyearlen / lno1wkst:

if lno1wkst >= 4:
    lno1wkst = 0
    lwyearlen = lyearlen+(lyearweekday-rr._wkst) % 7
else:
    lwyearlen = lyearlen-lno1wkst
ldiv, lmod = divmod(lwyearlen, 7)
lnumweeks = ldiv+lmod//4

Validation

• New regression test testYearlyByWeekNoLastWeekOfPrevYear (fails before, passes after).
• A systematic cross-check of byweekno=N against Python's own date.isocalendar() over every week of 1998-2034 had mismatches only in 2011 and 2022 before this change and zero after.
• The full tests/test_rrule.py suite passes: 562 passed, 1 xfailed (the 561 pre-existing tests plus the new regression test). The documented testYearlyByWeekNoAndWeekDay53 still returns its exact expected values.

Includes an AUTHORS.md entry and a changelog.d news fragment.