fix: name the trip on a shopping row bought only on a later trip

[guplem]

Summary

Closes #73

A product that only later shopping trips buy now names the trip on the shopping page, for example "Buy 2 pieces Saturday 17 Oct" (the shopping day) instead of a plain "Buy 2 pieces". Before, the trip appeared only in the copied list and the PDF, so the user could buy a perishable product too early.

• Standalone rows (_tripPurchasesForProduct, shopping_ingredient.dart): when one trip buys the product, the row names the earliest trip that holds the ingredient. When that is the first trip of the plan, the row stays a plain "Buy X". The 2+ trip split is unchanged.
• Equivalent-product groups (for example two pizza flavors of one size): each member row names the earliest trip that holds the ingredient, when that is not the first trip. It keeps its own one-of-each count.
• Row (ShoppingProductRow): the single-total line adds the trip label from shoppingTripLabel, the same function that the 2+ lines, the copy and the PDF use. The under-buy reduction still applies.
• Known difference: when the rounding of the split gives an early trip 0 units, the row names that earlier trip, but the copied list puts the product under the later one. This is on purpose (safer to buy early) and recorded in ADR 0014. The split fix is Rounding of the trip split can put a week-0 need on a later trip #75.
• Docs: ADR 0014 and the AGENTS.md test count (1066 tests).

Test plan

• 13 new tests, each proven by a mutation of the production line; 3 weak textContaining checks replaced by exact checks
• Format check, flutter analyze, full suite (1066 tests) pass locally
• Manual check: open the shopping page of a multi-week menu and confirm a later-trip-only product names its trip

🤖 Generated with Claude Code

[guplem]

Good, focused fix. The single later-trip line reuses shoppingTripLabel through the new _tripWhen helper, so the row, the 2+ trip lines, the copy and the PDF name the trip the same way. The under-buy case is handled, and each new test asserts the full text and was proven by a mutation.

Two gaps keep the fix from being complete:

• [Required] Equivalent-product groups still show a plain total (menu_management/lib/shopping/shopping_ingredient.dart:487, tripPurchases: isCombinedGroup ? const [] : _tripPurchasesForProduct(product),). Two pizza flavors of one size form such a group. When only a later trip buys the ingredient, each member row still reads "Buy 1 pack", which is the #73 symptom. The old reason for the empty list (a 2+ split would show wrong solo counts) does not apply to a label that only names the trip. Fix: when every purchase of the group falls on one later trip, pass a one-entry list to each visible member, and add a widget test.
• [Required] Rounding can name a trip that is too late (inline comment on line 292).

Smaller items:

• [Suggestion] textContaining("week") in test/shopping_ingredient_test.dart:232, :249 and :348 can never fail, because the labels write "Week" with a capital W. These tests now guard the "first trip alone" rule. Replace them with exact find.text(...) checks of the wrong outputs.
• [Suggestion] Add ingredient-level tests for: a start date (real date label through _tripPurchasesForProduct), owned stock that lowers the single later-trip count, and parity with the copied list's trip section.
• [Suggestion] adr/0014-multi-trip-shopping-planner.md:77 says the on-screen list shows no trip feedback besides the banner. That is out of date since #25 and more so now.
• Follow-up candidate: an ingredient with no product shows only "Need: X" and never names a trip.

[guplem]

Round 2. Both Required items from round 1 are fixed: equivalent-product rows name a single later trip, and the single-trip row names the earliest trip that holds the ingredient. The weak textContaining("week") checks are now exact, and ADR 0014 records the per-row trip label.

• [Suggestion] Equivalent group bought only on 2+ later trips still reads "buy now" (shopping_ingredient.dart, groupTripPurchases.length == 1 ? groupTripPurchases.single : null). Trips week 0 (rice only), week 1 and week 2 (pizza): each member row shows a plain "Buy 1 pack". Fix: name the earliest trip that holds the ingredient whenever it is not the first trip.
• [Suggestion] ADR 0014 says the copy always equals the on-screen list (line 75). In the rounding case the row names an earlier trip than the copy section, on purpose. Say so, so nobody "fixes" one side back. The split itself is tracked in #75.
• [Nitpick] if (purchases.length != 1) return purchases; // The split weighs ...: move the comment above the firstWhere that it explains.
• [Nitpick] isFirstTrip: groupSingleTrip.isFirstTrip can only be false; write isFirstTrip: false as the standalone branch does.
• [Nitpick] groupSingleTrip != null && packsToBuy > 0: the row shows "Covered" for 0 packs before it reads tripPurchases, so the check has no effect.
• [Nitpick] The owned-stock test uses remainingGrams: 300 and a trip item of 300, so a count from the raw trip amount also passes. Use a different trip amount (for example 900).