add period comparison deltas to analytics stat cards

Each stat card now shows the percentage change vs the equivalent
previous period (e.g. 30d compares last 30 days vs 30 days before).
Handles zero-baseline with "new" label and caps extreme deltas at
>999%. Seed data extended to 2 years for meaningful 12m comparisons.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

priv/repo/seeds/analytics.exs

@@ -1,8 +1,8 @@
-# Generates realistic analytics demo data spanning 12 months.
+# Generates realistic analytics demo data spanning 2 years.
 #
 #     mix run priv/repo/seeds/analytics.exs
 #
-# Clears existing analytics events first, then creates ~50k events with
+# Clears existing analytics events first, then creates ~90k events with
 # realistic traffic patterns, referrers, device mix, and e-commerce funnel.
 
 alias Berrypod.Repo
@@ -13,9 +13,9 @@ alias Berrypod.Analytics.Event
 # How many unique "visitors" to simulate per day (base — actual varies by day)
 base_daily_visitors = 40
 
-# Date range: 12 months back from today
+# Date range: 2 years back from today
 end_date = Date.utc_today()
-start_date = Date.add(end_date, -364)
+start_date = Date.add(end_date, -729)
 
 # ── Reference data ──
 
@@ -150,8 +150,8 @@ end
 # Monthly growth curve — traffic grows over time (new shop ramping up)
 month_multiplier = fn date ->
   months_ago = Date.diff(end_date, date) / 30.0
-  # Start at 0.3x and grow to 1.0x over the year
-  max(0.3, 1.0 - months_ago * 0.06)
+  # Start at 0.2x two years ago and grow to 1.0x now
+  max(0.2, 1.0 - months_ago * 0.035)
 end
 
 # Seasonal bumps (Nov-Dec holiday shopping, Jan sale)