perf(transform): optimize dim_locations spatial joins via IEJoin + country filters

Replace ABS() bbox predicates with BETWEEN in all three spatial CTEs
(nearest_padel, padel_local, tennis_nearby). BETWEEN enables DuckDB's
IEJoin (interval join) which is O((N+M) log M) vs the previous O(N×M)
nested-loop cross-join.

Add country pre-filters to restrict the left side from ~140K global
locations to ~20K rows for padel/tennis CTEs (~8 countries each).

Expected: ~50-200x speedup on the spatial CTE portion of the model.

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

transform/sqlmesh_padelnomics/models/foundation/dim_locations.sql

@@ -19,8 +19,10 @@
 --   4. Country-level income       (global fallback from stg_income / ilc_di03)
 --
 -- Distance calculations use ST_Distance_Sphere (DuckDB spatial extension).
--- A bounding-box pre-filter (~0.5°, ≈55km) reduces the cross-join before the
--- exact sphere distance is computed.
+-- Spatial joins use BETWEEN predicates (not ABS()) to enable DuckDB's IEJoin
+-- (interval join) optimization: O((N+M) log M) vs O(N×M) nested-loop.
+-- Country pre-filters restrict the left side to ~20K rows for padel/tennis CTEs
+-- (~8 countries each), down from ~140K global locations.
 
 MODEL (
   name foundation.dim_locations,
@@ -147,6 +149,8 @@ padel_courts AS (
   WHERE lat IS NOT NULL AND lon IS NOT NULL
 ),
 -- Nearest padel court distance per location (bbox pre-filter → exact sphere distance)
+-- BETWEEN enables DuckDB IEJoin (O((N+M) log M)) vs ABS() nested-loop (O(N×M)).
+-- Country pre-filter reduces left side from ~140K to ~20K rows (padel is ~8 countries).
 nearest_padel AS (
   SELECT
     l.geoname_id,
@@ -158,9 +162,12 @@ nearest_padel AS (
     ) AS nearest_padel_court_km
   FROM locations l
   JOIN padel_courts p
-    -- ~55km bounding box pre-filter to limit cross-join before sphere calc
-    ON ABS(l.lat - p.lat) < 0.5
-    AND ABS(l.lon - p.lon) < 0.5
+    -- ~55km bounding box pre-filter; BETWEEN triggers IEJoin optimization
+    ON l.lat BETWEEN p.lat - 0.5 AND p.lat + 0.5
+    AND l.lon BETWEEN p.lon - 0.5 AND p.lon + 0.5
+  WHERE l.country_code IN (
+    SELECT DISTINCT country_code FROM padel_courts WHERE country_code IS NOT NULL
+  )
   GROUP BY l.geoname_id
 ),
 -- Padel venues within 5km of each location (counts as "local padel supply")
@@ -170,24 +177,35 @@ padel_local AS (
     COUNT(*) AS padel_venue_count
   FROM locations l
   JOIN padel_courts p
-    ON ABS(l.lat - p.lat) < 0.05    -- ~5km bbox pre-filter
-    AND ABS(l.lon - p.lon) < 0.05
-  WHERE ST_Distance_Sphere(
+    -- ~5km bbox pre-filter; BETWEEN triggers IEJoin optimization
+    ON l.lat BETWEEN p.lat - 0.05 AND p.lat + 0.05
+    AND l.lon BETWEEN p.lon - 0.05 AND p.lon + 0.05
+  WHERE l.country_code IN (
+    SELECT DISTINCT country_code FROM padel_courts WHERE country_code IS NOT NULL
+  )
+    AND ST_Distance_Sphere(
       ST_Point(l.lon, l.lat),
       ST_Point(p.lon, p.lat)
     ) / 1000.0 <= 5.0
   GROUP BY l.geoname_id
 ),
 -- Tennis courts within 25km of each location (sports culture proxy)
+-- Country pre-filter reduces left side from ~140K to ~20K rows (tennis courts are European only).
 tennis_nearby AS (
   SELECT
     l.geoname_id,
     COUNT(*) AS tennis_courts_within_25km
   FROM locations l
   JOIN staging.stg_tennis_courts t
-    ON ABS(l.lat - t.lat) < 0.23   -- ~25km bbox pre-filter
-    AND ABS(l.lon - t.lon) < 0.23
-  WHERE ST_Distance_Sphere(
+    -- ~25km bbox pre-filter; BETWEEN triggers IEJoin optimization
+    ON l.lat BETWEEN t.lat - 0.23 AND t.lat + 0.23
+    AND l.lon BETWEEN t.lon - 0.23 AND t.lon + 0.23
+  WHERE l.country_code IN (
+    SELECT DISTINCT country_code
+    FROM staging.stg_tennis_courts
+    WHERE country_code IS NOT NULL
+  )
+    AND ST_Distance_Sphere(
       ST_Point(l.lon, l.lat),
       ST_Point(t.lon, t.lat)
     ) / 1000.0 <= 25.0