feat: Implement Phase 4 dbt model restructuring

Create neighbourhood-centric dbt transformation layer:

Staging (5 models):
- stg_toronto__neighbourhoods - Neighbourhood dimension
- stg_toronto__census - Census demographics
- stg_toronto__crime - Crime statistics
- stg_toronto__amenities - Amenity counts
- stg_cmhc__zone_crosswalk - Zone-to-neighbourhood weights

Intermediate (5 models):
- int_neighbourhood__demographics - Combined census with quintiles
- int_neighbourhood__housing - Housing + affordability indicators
- int_neighbourhood__crime_summary - Aggregated crime with YoY
- int_neighbourhood__amenity_scores - Per-capita amenity metrics
- int_rentals__neighbourhood_allocated - CMHC via area weights

Marts (5 models):
- mart_neighbourhood_overview - Composite livability score
- mart_neighbourhood_housing - Affordability index
- mart_neighbourhood_safety - Crime rates per 100K
- mart_neighbourhood_demographics - Income/age indices
- mart_neighbourhood_amenities - Amenity index

Closes #60, #61, #62, #63

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

dbt/models/marts/mart_neighbourhood_safety.sql

@@ -0,0 +1,78 @@
+-- Mart: Neighbourhood Safety Analysis
+-- Dashboard Tab: Safety
+-- Grain: One row per neighbourhood per year
+
+with crime as (
+    select * from {{ ref('int_neighbourhood__crime_summary') }}
+),
+
+-- City-wide averages for comparison
+city_avg as (
+    select
+        year,
+        avg(crime_rate_per_100k) as city_avg_crime_rate,
+        avg(assault_count) as city_avg_assault,
+        avg(auto_theft_count) as city_avg_auto_theft,
+        avg(break_enter_count) as city_avg_break_enter
+    from crime
+    group by year
+),
+
+final as (
+    select
+        c.neighbourhood_id,
+        c.neighbourhood_name,
+        c.geometry,
+        c.population,
+        c.year,
+
+        -- Total crime
+        c.total_incidents,
+        c.crime_rate_per_100k,
+        c.yoy_change_pct as crime_yoy_change_pct,
+
+        -- Crime breakdown
+        c.assault_count,
+        c.auto_theft_count,
+        c.break_enter_count,
+        c.robbery_count,
+        c.theft_over_count,
+        c.homicide_count,
+
+        -- Per 100K rates by type
+        case when c.population > 0
+            then round(c.assault_count::numeric / c.population * 100000, 2)
+            else null
+        end as assault_rate_per_100k,
+
+        case when c.population > 0
+            then round(c.auto_theft_count::numeric / c.population * 100000, 2)
+            else null
+        end as auto_theft_rate_per_100k,
+
+        case when c.population > 0
+            then round(c.break_enter_count::numeric / c.population * 100000, 2)
+            else null
+        end as break_enter_rate_per_100k,
+
+        -- Comparison to city average
+        round(ca.city_avg_crime_rate::numeric, 2) as city_avg_crime_rate,
+
+        -- Crime index (100 = city average)
+        case
+            when ca.city_avg_crime_rate > 0
+            then round(c.crime_rate_per_100k / ca.city_avg_crime_rate * 100, 1)
+            else null
+        end as crime_index,
+
+        -- Safety tier based on crime rate percentile
+        ntile(5) over (
+            partition by c.year
+            order by c.crime_rate_per_100k desc
+        ) as safety_tier
+
+    from crime c
+    left join city_avg ca on c.year = ca.year
+)
+
+select * from final