fire-planner: life-event spending bumps now reflected in fan + auto-

refresh on scenario edits

Two fixes for the user's report that adding a £100k life-event spend
didn't change the chart:

Engine (simulator.py)
- New `extra_outflows` param. cashflow_adjustments still drains the
  portfolio at start-of-year as before, but the simulator now ALSO
  records the spending in `withdrawal_hist[p, y]` so the chart's red
  median-withdrawal trace shows the bump. Without this, the £100k
  silently came out of the portfolio but the user-facing withdrawal
  trace stayed at the strategy's flat 4% draw.
- simulate.py wires extra_outflows = essential + discretionary
  category outflows from life events.

UX (ScenarioDetail.tsx)
- New auto-refresh: when life events / income streams / flex rules
  change for a scenario, the page fires `/simulate` automatically
  with 2,000 paths and uses the result as the primary fan/year-stats
  source. The persisted MC run is only consulted as a fallback for
  scenarios with no overrides.
- Fan chart title gains a "live preview · Xs · Ny" pill while a sim
  is current, and "re-running…" while a fresh one is in flight.
- Removed the now-redundant "Live preview run" duplicate card lower
  down — the main chart IS the live preview.
- Year-stats badge row reads from sim.data when available so changes
  propagate immediately to NW / Δ NW / Spending / Taxes.

247 pytest pass (+1 new); mypy + ruff clean; frontend typecheck/test/
build green.

tests/test_simulator_events.py

@@ -2,6 +2,7 @@
 from __future__ import annotations
 
 import numpy as np
+import pytest
 
 from fire_planner.glide_path import static
 from fire_planner.life_events import EventInput, events_to_cashflow_array
@@ -30,6 +31,27 @@ def test_no_adjustments_matches_baseline() -> None:
     np.testing.assert_allclose(base.portfolio_real, with_zero.portfolio_real)
 
 
+def test_extra_outflows_show_up_in_withdrawal_trace() -> None:
+    """A £100k spending bump in years 5-10 should be visible on the
+    withdrawal trace — not just silently drained from the portfolio."""
+    kwargs = _baseline_kwargs()
+    adj = np.zeros(25, dtype=np.float64)
+    extras = np.zeros(25, dtype=np.float64)
+    adj[5:11] = -100_000.0  # drains the portfolio
+    extras[5:11] = 100_000.0  # surfaces on the chart
+
+    base = simulate(**kwargs)  # type: ignore[arg-type]
+    bumped = simulate(**kwargs, cashflow_adjustments=adj, extra_outflows=extras)  # type: ignore[arg-type]
+
+    # Year 0–4 unchanged (no extra outflow)
+    np.testing.assert_allclose(base.withdrawal_real[:, :5], bumped.withdrawal_real[:, :5])
+    # Years 5–10 should be ~100k higher than baseline (clipped only when
+    # the portfolio was already drained — checked by spot-test).
+    assert (bumped.withdrawal_real[:, 5:11] > base.withdrawal_real[:, 5:11]).all()
+    # Year 5 specifically: strategy w (~40k) + 100k extra ≈ 140k.
+    assert bumped.withdrawal_real[0, 5] == pytest.approx(140_000.0, rel=0.05)
+
+
 def test_one_time_inheritance_lifts_portfolio() -> None:
     kwargs = _baseline_kwargs()
     adj = events_to_cashflow_array(