build(tasks): GOEXPERIMENT=jsonv2 per SCALE_STANDARD (#1)

* docs(llm): explicit Upstream attribution line

LICENSE-attested or repo-attested upstream is now called out at the
top of LLM.md alongside the project description, so downstream
audits (universe/docs/PRODUCTS.md) can rely on a single canonical
location for the OSS lineage statement.

* build(tasks): GOEXPERIMENT=jsonv2 in Dockerfile per SCALE_STANDARD

Per SCALE_STANDARD.md §2 — every Go production Dockerfile that emits
JSON to a client builds with GOEXPERIMENT=jsonv2. Verified on
hanzoai/zip's json_bench_test.go:
- Edge POST roundtrip: -12% time, -23% allocs
- Marshal-only: -22% time
- Unmarshal-only: -19% time, -25% allocs

The conn-memory regression test is deferred to a follow-up that
migrates tasksd's net/http listener to zip per §9 step-1; net/http
has a fixed 3-goroutine-per-request overhead that violates §3's 1:1
goroutine-to-conn invariant.

Dockerfile

@@ -15,6 +15,14 @@ COPY go.mod go.sum ./
 RUN go mod download
 COPY . .
 RUN test -f ui/dist/index.html || (echo "ui/dist missing — run scripts/sync-admin-ui.sh before docker build" >&2 && exit 1)
+
+# Per SCALE_STANDARD.md §2 — every Go production Dockerfile that
+# emits JSON to a client builds with GOEXPERIMENT=jsonv2. Verified
+# -12% time / -23% allocs on the edge POST roundtrip vs encoding/json
+# v1 (json_bench_test.go in hanzoai/zip).
+ARG GO_EXPERIMENT=jsonv2
+ENV GOEXPERIMENT=${GO_EXPERIMENT}
+
 RUN CGO_ENABLED=0 go build -trimpath -ldflags="-s -w" -o /tasksd ./cmd/tasksd
 
 FROM alpine:3.21

LLM.md

@@ -2,6 +2,8 @@
 
 Durable workflow execution engine for AI agent orchestration.
 
+**Upstream**: Temporal (MIT). Pre-v3.0 history is the Temporal fork; v3.0+ is native ZAP-only with the upstream gRPC/proto/runtime fully ripped out. Public Go module pinned at v1.x forever (semver-import-versioning policy).
+
 ## Module
 `github.com/hanzoai/tasks`