Metric: price impact as a function of trade size

Written to the standard set by #33 — read that issue first for the expected level of detail and the acceptance criteria every check shares. ## What Measure **price impact as a function of size** — how much worse the effective rate gets as the trade grows — and report it as a curve, not a single number. ## Why it matters The ladder already shows this happening: USDC→NGNC loses 25.02% at 0.1 USDC and 97.68% at 5000. But nothing names the *shape*. A corridor with a flat curve and a bad floor is a different problem from one with a good floor and a cliff — the first is priced badly, the second is thin. Separating the two is what tells a user whether their size is the problem or the corridor is. ## Which layer **Layer 2 — deterministic calculation** over the ladder's layer 1 facts. ## This is a metric Returns `MetricResult` per size, not a verdict. See "Checks and metrics are separate shapes" in [docs/checks.md](https://github.com/Wayfare-labs/wayfare/blob/main/docs/checks.md). **Do not add a threshold** — deciding what impact is unacceptable is maintainer-owned. ## Exactly where the code goes - **New file:** `checks/metric_price_impact.go` - Tests appended to `checks/checks_test.go` - `dex.MeasureSlippage` in `dex/dex.go` already compares a full size against a probe — reuse it rather than reimplementing ## Where the data comes from `dex.Client.StrictSendPaths` across `route.DefaultSizes`, which the ladder already walks. ## What "unmeasurable" means here - Fewer than two priced sizes → `Determined: false`. Impact is a comparison; one point is not a curve - No market at any size → undetermined, **not zero impact**. Zero impact would say "this corridor is perfectly liquid", the opposite of the truth ## Acceptance criteria - [ ] `MetricResult` per size, with `UnitPercent` - [ ] `decimal.Decimal` throughout; no float64 - [ ] Returns UNABLE-TO-DETERMINE (not zero) with fewer than two priced sizes - [ ] Reuses `dex.MeasureSlippage` rather than recomputing - [ ] Evidence names the sizes compared - [ ] Table tests from `testdata/snapshots` — NGNC for a real curve, KESC for the no-market case - [ ] Negative test: a corridor with one or zero priced sizes - [ ] No new third-party dependencies - [ ] gofmt, vet, test -race, golangci-lint clean - [ ] Adds no threshold turning the metric into a verdict ## Out of scope Nothing in `dex/` may change. Nothing in `route/`, `sep38/`, `runstore/runstore.go`. ## How to verify locally ```bash make offline-test go test ./checks/ -run TestPriceImpact -v ```