…ation

Fixes review comment: VTIMEZONE was hardcoding US DST transition rules
(2nd Sunday of March, 1st Sunday of November) for all DST timezones.

This caused incorrect UTC offsets for:
- European timezones (transition in late March/late October)
- Southern Hemisphere timezones (transitions in opposite months)
- Many other IANA zones with different transition patterns

Changes:
- Add findDSTTransition() helper: binary-searches the actual transition
  moment in 1970 by detecting when utcOffset() changes
- Add formatTransitionDtstart(): formats transition as iCal DTSTART
- Add getBydayRule(): computes nth-weekday RRULE BYDAY pattern from date
- buildVTimezone() now uses these helpers instead of hardcoded US dates
- Add tests for Europe/London (March/October) and Australia/Sydney
  (Southern Hemisphere reverse DST) to prevent regression

…eling

Bug 1 (DTSTART 9-digit date): formatTransitionDtstart used template literal
`19700${month}${day}` where month is already padStart(2,'0') (e.g. '03').
This produced 9-digit dates like '197000308' instead of valid 8-digit '19700308'.
Fix: remove the spurious '0' → `1970${month}${day}`.

Bug 2 (Southern Hemisphere DAYLIGHT/STANDARD inversion): The previous code
used isNorthernHemisphereStyle to swap offset values within fixed DAYLIGHT/
STANDARD blocks. For Southern Hemisphere zones (e.g. Australia/Sydney) where
January is summer/daylight:
- springTransition (Jan→Jul) finds April — going from AEDT→AEST (DST END)
- fallTransition (Jul→Jan) finds October — going from AEST→AEDT (DST START)
But the code always emitted BEGIN:DAYLIGHT for springTransition, incorrectly
labeling April (DST end) as daylight and October (DST start) as standard.

Fix: determine the semantic meaning of each transition by comparing UTC offsets:
- springIsDaylight = (summerUtcOffset > winterUtcOffset)
  - NH (e.g. America/Chicago): Jul(+summer) > Jan(+winter) → spring = DAYLIGHT ✓
  - SH (e.g. Australia/Sydney): Jul(+winter) < Jan(+summer) → spring = STANDARD ✓
- Also compute trueStandardOffset/trueDaylightOffset based on which period
  actually has the higher UTC offset (daylight always has higher offset).

Test additions:
- DTSTART 8-digit regression test for America/New_York
- Australia/Sydney now asserts BYMONTH=10 in DAYLIGHT and BYMONTH=4 in STANDARD,
  with correct TZOFFSETFROM/TZOFFSETTO (+1000↔+1100)

cubic-dev-ai P2: VTIMEZONE RRULEs derived from 1970 transitions can produce
outdated BYDAY/BYMONTH for zones that changed DST rules after 1970.

The US changed DST rules in 2007 (Energy Policy Act): spring transition
moved from 1st Sunday in April to 2nd Sunday in March. Using 1970 to find
transitions would yield BYMONTH=4;BYDAY=1SU (pre-2007 rules) instead of
the correct BYMONTH=3;BYDAY=2SU (current rules).

Fix: pass the event's actual year to findDSTTransition() instead of 1970,
so BYDAY/BYMONTH always reflects the rules in effect for the event's year.
formatTransitionDtstart() now uses the actual transition year for DTSTART.

Test additions:
- Update existing 8-digit DTSTART test comments to reflect new year format
- Add regression test verifying America/New_York uses BYMONTH=3 (March, post-2007)
  not BYMONTH=4 (April, pre-2007/1970 rules)