feat(pipeline): add Meltano + dbt + Airflow ELT pipeline scaffold

Replaces the hand-rolled integrator with a production-grade ELT pipeline
using Meltano (Singer taps), dbt Core (medallion architecture), and
Apache Airflow (orchestration). Adds Typesense for search and PostGIS
for geospatial queries.

- 6 custom Singer taps (GIAS, EES, Ofsted, Parent View, FBIT, IDACI)
- dbt project: 12 staging, 5 intermediate, 12 mart models
- 3 Airflow DAGs (daily/monthly/annual schedules)
- Typesense sync + batch geocoding scripts
- docker-compose: add Airflow, Typesense; upgrade to PostGIS
- Portainer stack definition matching live deployment topology

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

pipeline/transform/models/marts/fact_deprivation.sql

@@ -0,0 +1,22 @@
+-- Mart: Deprivation index — one row per URN
+-- Joins school postcode → LSOA → IDACI score
+
+with school_postcodes as (
+    select
+        urn,
+        postcode
+    from {{ ref('stg_gias_establishments') }}
+    where status = 'Open'
+      and postcode is not null
+)
+
+-- Note: The join between postcode and LSOA requires a postcode-to-LSOA
+-- lookup table. This will be populated by the geocode script or a seed.
+-- For now, this model serves as a placeholder that will be completed
+-- once the IDACI tap provides the postcode→LSOA mapping.
+
+select
+    i.lsoa_code,
+    i.idaci_score,
+    i.idaci_decile
+from {{ ref('stg_idaci') }} i