feat: add loaders and dbt models for Toronto housing data

Sprint 4 implementation:

Loaders:
- base.py: Session management, bulk insert, upsert utilities
- dimensions.py: Load time, district, zone, neighbourhood, policy dimensions
- trreb.py: Load TRREB purchase data to fact_purchases
- cmhc.py: Load CMHC rental data to fact_rentals

dbt Project:
- Project configuration (dbt_project.yml, packages.yml)
- Staging models for all fact and dimension tables
- Intermediate models with dimension enrichment
- Marts: purchase analysis, rental analysis, market summary

Closes #16

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

dbt/models/marts/mart_toronto_purchases.sql

@@ -0,0 +1,79 @@
+-- Mart: Toronto Purchase Market Analysis
+-- Final analytical table for purchase/sales data visualization
+-- Grain: One row per district per month
+
+with purchases as (
+    select * from {{ ref('int_purchases__monthly') }}
+),
+
+-- Add year-over-year calculations
+with_yoy as (
+    select
+        p.*,
+
+        -- Previous year same month values
+        lag(p.avg_price, 12) over (
+            partition by p.district_code
+            order by p.date_key
+        ) as avg_price_prev_year,
+
+        lag(p.sales_count, 12) over (
+            partition by p.district_code
+            order by p.date_key
+        ) as sales_count_prev_year,
+
+        lag(p.median_price, 12) over (
+            partition by p.district_code
+            order by p.date_key
+        ) as median_price_prev_year
+
+    from purchases p
+),
+
+final as (
+    select
+        purchase_id,
+        date_key,
+        full_date,
+        year,
+        month,
+        quarter,
+        month_name,
+        district_key,
+        district_code,
+        district_name,
+        area_type,
+        sales_count,
+        dollar_volume,
+        avg_price,
+        median_price,
+        new_listings,
+        active_listings,
+        days_on_market,
+        sale_to_list_ratio,
+        absorption_rate,
+        months_of_inventory,
+
+        -- Year-over-year changes
+        case
+            when avg_price_prev_year > 0
+            then round(((avg_price - avg_price_prev_year) / avg_price_prev_year) * 100, 2)
+            else null
+        end as avg_price_yoy_pct,
+
+        case
+            when sales_count_prev_year > 0
+            then round(((sales_count - sales_count_prev_year)::numeric / sales_count_prev_year) * 100, 2)
+            else null
+        end as sales_count_yoy_pct,
+
+        case
+            when median_price_prev_year > 0
+            then round(((median_price - median_price_prev_year) / median_price_prev_year) * 100, 2)
+            else null
+        end as median_price_yoy_pct
+
+    from with_yoy
+)
+
+select * from final