Enter owned quantity per product and auto-calculate the global owned amount

[guplem]

Summary

Closes #24

Before, the shopping list took one "owned" amount per ingredient in a single unit. Now the user can say how many they own of each product, and the ingredient's global owned amount is computed automatically in grams/ml.

• Each product row has its own "Owned" count input.
• Global owned = sum over products of count x pack quantity, converted to the required unit through the ingredient's existing conversions (toGrams/fromGrams, bridging by density or grams-per-piece).
• Remaining-to-buy and trip planning use this computed amount, through one shared resolver (OwnedStock.amountInUnit), so the on-screen "Need" and the copied amounts stay identical.
• Ingredients with no products keep the single owned input plus unit dropdown (fallback preserved).

Test plan

• flutter analyze passes
• flutter test test/ passes (633 tests)
• Owning several products of different pack sizes sums to the right grams and lowers the amount to buy
• A no-products ingredient still shows one owned input

🤖 Generated with Claude Code

[guplem]

Reviewed by an automated pass (deep review + a fix round). One real gap was found and fixed on the branch; the change is now correct and ready for a human check.

What is done well

• Clean conversion logic: the global owned amount sums each product's count x pack quantity and converts through the ingredient's real toGrams/fromGrams (bridging by density or grams-per-piece). No double-counting; per-product indices are stable and read back through bounds-guarded lookups.
• Page and copy use one shared resolver (OwnedStock.amountInUnit), so the on-screen Need and the copied amounts stay identical, including the multi-trip planner. Good tests on real summed grams (3x500 + 5x250 = 2750), pieces-via-grams-per-piece, and the planner buy amount.
• No-products fallback preserved.

Gap found and fixed on this branch (was a silent loss of function)

• An ingredient that has products but where no product unit matches any recipe unit (e.g. a pieces-only product used by a grams recipe) rendered no product rows AND hid the header input, so the user could enter no owned amount at all. Fixed: a shared usesPerProductOwnedInputs(...) helper now decides both the per-product rows and the header fallback from the same condition, so the header input returns whenever no per-product input would show. A widget test proves entering owned then reduces the on-screen Need. Also hardened the row controller with didUpdateWidget + a ValueKey. Full suite green (635 tests).

Verdict: approved pending human review. CI analyze-and-test gate must be green before merge.

Note for the reviewer: this PR and #39 both change owned_amount.dart and multi_trip_planner.dart, so expect a merge conflict when landing the second one.

[guplem]

Rebased onto the updated main (which now has #34, #29, #28) and reconciled. CI is green, the PR is mergeable, and a fresh correctness review of the reconciliation passed.

Reconciliation (verified, this was the hardest merge)

• Per-product owned now flows through #34's single shared owned-pool. OwnedStockConsumer/computeRemainingQuantities were changed to take an OwnedStock; both the single-amount and per-product shapes resolve their grams total via OwnedStock.amountInUnit(ingredient, grams), which seeds one pool that is drawn down once across units. No per-unit double subtraction was reintroduced (the #34 bug guarantee survives).
• Page and planner consume owned through the same OwnedStockConsumer via one _ownedStockFor helper, so they cannot disagree. The old per-unit subtraction loops were removed.
• Per-product grams total = sum of count x pack quantity converted once via toGrams/fromGrams. Product-index keying is bounds-guarded.
• Card keeps the per-product owned inputs (#24), the under-buy chip (#29), and the freeze note (#28) together.
• No asserted test value was weakened; the computeRemainingQuantities signature change was applied consistently to all call sites. A new test proves per-product owned is subtracted once across a pieces+grams need (pieces 0, grams 50). Full suite passes together: 661 tests. ADR 0014 reconciled.

Verdict: approved pending human review. CI green and mergeable.

Note: this and #34 both rewrote owned_amount.dart (#34 is already merged), so this PR carries a merge commit reconciling them; the diff vs current main is the net per-product-owned addition.

[guplem]

Re-merged after #37 landed on main. CI is green and the PR is mergeable again.

Reconciliation with #37 (verified)

• Only shopping_ingredient.dart conflicted (the product-row build). #37's _buildProductRows grouping (combined "buy one of each" for equivalent products, cycled shares, "and"/"or" dividers, zero-share skipping) is kept, and each rendered row now also carries #40's per-product owned input: ownedCount + onOwnedCountChanged keyed by the product's true index, plus a ValueKey. So each member of a combined equivalent group still shows its own "Owned" input.
• The header-fallback owned input stays mutually exclusive with per-product inputs (!usesPerProductOwnedInputs).
• Copy needed no manual edit: it still computes owned-adjusted remaining via computeRemainingQuantities (#40) and then distributes one-of-each across equivalent products via buildIngredientCopyLines (#37).
• Test change (justified, values unchanged): two auto-fill tests were reframed to the header-fallback scenario because per-product inputs now hide the header auto-fill button for grams-product + grams-recipe; the summing logic and expected values (3.0, 1.0) are unchanged. Full suite passes together: 683 tests.

Verdict: approved pending human review. CI green and mergeable.

Note: #42 is also conflicting now for the same reason (it shares the card/copy code with #37) - it needs the same reconciliation before it can merge.