Keep a repeating week on its weekday when the planner picks the dates

A yearly-repeating period kept the same calendar date instead of the same
weekday: 2026-10-05..10-11 (Mon-Sun) recurred as 2027-10-05..10-11 (Tue-Mon)
rather than 2027-10-04..10-10. The snap added in 224fe7a only ever covered
window_mode == 'fixed'; 'flexible' wishes were unaligned at both stages.

project_window now snaps both modes. A flexible bound is still a bounding
range, but it keeps its weekday alignment year over year with its span held
fixed, so the search space never shrinks and a bound that is exactly as long
as the target (its only candidate) repeats week-shaped.

_rank_candidate gains a Monday-start term between cost and start, so equal-cost
candidates break on week alignment instead of "earliest". It sits after cost
deliberately: a cheaper non-Monday start still wins, because red days outrank
alignment.

_reuse_already_funded_window would otherwise freeze a pre-existing misaligned
placement forever, so it now releases a year whose reuse winner is misaligned
when a fresh search ranks strictly better -- never when one of its days is
pinned or excluded, so a deliberate per-day override always survives.

Migration 008 realigns flexible series already materialised under the old rule;
projection runs once, at create/update, so nothing repairs them on read.

Alongside this, the planner now says when it moved a period to catch red days.
WindowOutcome carries shift_gain_days (the saving over the naive bound-start
placement) and the calendar toasts it, once, on the sync that decided it.
A fixed window is never auto-moved -- its "exact requested placement,
unconditionally" guarantee stands -- but its shift suggestion is no longer
hidden behind a shortfall, so a fully-funded one can surface a cheaper week too.

Editing a period in the popover now also sends window_mode: 'fixed' with a
matching target. Preset chips create flexible wishes; dragging one down to an
exact week left it flexible with a week-long bound, which is the degenerate
single-candidate case that produced the reported dates.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DXPBfr1v7F71t6QcbHCaaT

backend/app/entities/leave_wish.py

@@ -79,28 +79,37 @@ def project_window(
     Nothing re-derives it at read time — `LeaveWish.window()` just reads the
     dates back off the row.
 
-    `window_mode == 'fixed'` wishes describe a week-based habit, not a literal
-    month/day: someone who is off Mon-Sun this year means Mon-Sun again next
-    year, even though the naive month/day anchor isn't a Monday every year
-    (the reported bug: 2026-10-05..10-11, Mon-Sun, naively recurred to
-    2027-10-05..10-11, which is Tue-Mon). So for any `year` other than
-    `anchor_year`, the naive `date(year, start_month, start_day)` is snapped
-    to the nearest date carrying the anchor's *start weekday* (ties resolve to
-    the earlier date — see `_shift_to_weekday`), and the end date is
-    re-derived by holding the anchor window's day-*span* fixed rather than
-    re-anchoring end_month/end_day. `anchor_year` itself is unshifted.
+    Both `window_mode`s describe a week-based habit, not a literal month/day:
+    someone who is off Mon-Sun this year means Mon-Sun again next year, even
+    though the naive month/day anchor isn't a Monday every year (the reported
+    bug: 2026-10-05..10-11, Mon-Sun, naively recurred to 2027-10-05..10-11,
+    which is Tue-Mon). So for any `year` other than `anchor_year`, the naive
+    `date(year, start_month, start_day)` is snapped to the nearest date
+    carrying the anchor's *start weekday* (ties resolve to the earlier date
+    — see `_shift_to_weekday`), and the end date is re-derived by holding the
+    anchor window's day-*span* fixed rather than re-anchoring
+    end_month/end_day. `anchor_year` itself is unshifted.
 
-    `window_mode == 'flexible'` wishes (§10.2) describe a BOUNDING RANGE the
-    allocator searches inside, not a placement — they keep the plain month/day
-    anchor with no weekday shift, every year. Each year's row therefore
-    carries that year's bounding range, and the per-year placement search
-    inside it still happens at plan time (`_Ledger.resolve_flexible_window`).
+    For `window_mode == 'fixed'` the span being held fixed IS the placement:
+    the result is the exact dates the family is home. For `window_mode ==
+    'flexible'` (§10.2) the result is still a BOUNDING RANGE the allocator
+    searches inside, not a placement — but the range itself must keep its
+    weekday alignment year over year, or a week-shaped bound (e.g. Mon-Sun)
+    degenerates into an arbitrary weekday every other year, which is exactly
+    as wrong for a search bound as for a fixed placement (a week-shaped bound
+    should stay week-shaped). The snap never changes the bound's *length* —
+    only `_shift_to_weekday`'s ±3-day slide of the start, with the same
+    day-span held fixed — so the search space `resolve_flexible_window`
+    explores never shrinks. `window_mode` is otherwise unused here; kept as a
+    parameter because both modes now take the identical snap path and every
+    caller already has the value at hand — removing it would just make call
+    sites pass a value nobody reads.
     """
     wraps = window_wraps_month_day(start_month, start_day, end_month, end_day)
     naive_start = clamped_date(year, start_month, start_day)
     naive_end = clamped_date(year + 1 if wraps else year, end_month, end_day)
 
-    if window_mode != "fixed" or year == anchor_year:
+    if year == anchor_year:
         return naive_start, naive_end
 
     anchor_start = clamped_date(anchor_year, start_month, start_day)
@@ -140,7 +149,12 @@ class LeaveWish:
     # are the exact placement, as above. 'flexible': the SAME four fields instead
     # describe the BOUNDING RANGE the allocator searches inside for the
     # best-costing placement (e.g. "vår" = 03-01..05-31) — target_weeks is still
-    # the wanted duration inside that range.
+    # the wanted duration inside that range. Both modes get the same
+    # per-year weekday snap at materialisation (`project_window`): a fixed
+    # window's exact placement keeps its weekday, AND a flexible window's
+    # bounding range keeps its weekday alignment (so a week-shaped bound
+    # stays week-shaped every year) while its span — and so its search
+    # space — stays exactly as wide as the anchor year's.
     window_mode: str = "fixed"  # 'fixed' | 'flexible'
 
     # Groups the per-year rows a repeating wish was materialised into. `None`