feat(api,ui): showcase pipeline — richer data + V1/V2 modeling foundation (PRP-38)

[w7-mgfcode]

Scope

PRP-38 — Showcase Rich Demo Control Center A: Data + V1/V2 Modeling Foundation
(file: PRPs/PRP-38-showcase-data-modeling-lifecycle.md)

Foundation slice of the four-PRP /showcase rich-demo epic
(PRPs/INITIAL/INITIAL-showcase-rich-demo-control-center.md). Turns the flat
11-step baseline-only /showcase into a phase-grouped, scenario-aware
operator demo that:

• Renders 6 phase cards (data / modeling / decision / verify / agent / cleanup) as a shadcn accordion that auto-expands the running phase.
• Adds a scenario picker (shadcn Select) with demo_minimal (default), showcase_rich (new), sparse.
• Adds a showcase_rich preset (5 stores × 15 products × 180 days) wired through SeederConfig.from_scenario.
• Adds data-phase enrichment steps (phase2_enrichment + historical_backfill) via two new /seeder/* endpoints.
• Adds ONE V2 prophet_like run in the modeling phase (registers with full artifacts/models/... artifact_uri — R1).
• Evolves backtest to feature-aware with bucket metrics (h_1_7/h_8_14/h_15_28/h_29_plus) shown inline.
• Per-step Inspect deep links for train, v2_train, register, backtest.

Non-goals (deferred to PRP-39/40/41)

• Champion-compat compare, stale-alias trigger, safer-Promote dialog, portfolio batch (PRP-39).
• Scenario simulate/save/compare, RAG corpus indexing, embedding-provider probe (PRP-40).
• Agent HITL, ops snapshot KPI strip, Inspect-Artifacts post-run panel, localStorage run history, Stop button, walkthrough docs polish (PRP-41).

Acceptance gates

• Task 1 (Contract Probe) committed at PRPs/ai_docs/prp-38-contract-probe-report.md with every cited field marked PRESENT.
• Backend _phase_table() and frontend PHASE_DEFS lockstep enforced by paired tests.
• /seeder/phase2-enrichment + /seeder/historical-activity route tests (2xx + RFC 7807 error path).
• ScenarioPreset.SHOWCASE_RICH non-NaN backtest WAPE invariant.
• step_v2_train writes artifact_uri = train_response["model_path"] (FULL path — R1) + runtime_info_extras.feature_frame_version=2.
• step_backtest on SHOWCASE_RICH posts include_baselines=true + feature-aware model_config_main; response's bucketed_aggregated_metrics non-empty.
• /showcase default still ≤ 90 s; scenario=showcase_rich ≤ 240 s soft / ≤ 300 s hard.
• All five validation gates green (ruff + ruff format + mypy + pyright + pytest unit+integration + migration-check); pnpm lint && pnpm tsc --noEmit -p tsconfig.app.json && pnpm test --run green; vertical-slice grep guards empty; no radix-ui barrel imports.

Branch: feat/showcase-38-data-modeling-lifecycle
PRP: PRPs/PRP-38-showcase-data-modeling-lifecycle.md
Epic INITIAL: PRPs/INITIAL/INITIAL-showcase-rich-demo-control-center.md

[w7-mgfcode]

Merged on dev via PR #310 — PRP-38 (showcase pipeline richer data + V1/V2 modeling foundation) is functionally complete. Closing as hygiene; the PRP-38 → 41 showcase epic landed end-to-end through PR #323 (511ff41).