fix(pipeline) : Make the soliguide addresses a bit more correct

They always contain the postcode and city name, which unfortunately
makes the BAN less reliable :'(

Example:

    curl https://api-adresse.data.gouv.fr/search/?q=311 Av lou Gabian, 83600 Fréjus 83600 Fréjus
    result_score=0.54

    curl https://api-adresse.data.gouv.fr/search/?q=311 Av lou Gabian, 83600 Fréjus
    result_score=0.73

A y regarder de plus près dans les sources, Soliguide a des addresses
contenant toujours code postal et commune. Une regexp pour y remédier !

   position__address                                                         after_regexp
   11 Rue Louis Apffel, 67000 Strasbourg                                   |  11 Rue Louis Apffel
    9 Rue Déserte, 67000 Strasbourg                                        |   9 Rue Déserte
    107 Ave Parmentier, 75011 Paris                                        | 107 Ave Parmentier
    2 Rue Bartisch, 67100 Strasbourg                                       |   2 Rue Bartisch
    1 Rue du Rempart, 67000 Strasbourg                                     |   1 Rue du Rempart
    Hôpital Civil de Strasbourg, 1 Pl. de l'Hôpital, 67000 Strasbourg      | Hôpital Civil de Strasbourg, 1 Pl. de l'Hôpital

pipeline/dbt/models/intermediate/sources/soliguide/int_soliguide__adresses.sql

@@ -4,17 +4,18 @@ WITH lieux AS (
 
 final AS (
     SELECT
-        lieu_id                          AS "id",
-        _di_source_id                    AS "source",
-        position__coordinates__x         AS "longitude",
-        position__coordinates__y         AS "latitude",
-        position__additional_information AS "complement_adresse",
-        position__city                   AS "commune",
-        position__address                AS "adresse",
-        position__postal_code            AS "code_postal",
+        lieu_id                                                  AS "id",
+        _di_source_id                                            AS "source",
+        position__coordinates__x                                 AS "longitude",
+        position__coordinates__y                                 AS "latitude",
+        position__additional_information                         AS "complement_adresse",
+        position__city                                           AS "commune",
+        REGEXP_REPLACE(position__address, ', \d\d\d\d\d.*$', '') AS "adresse",
+        position__postal_code                                    AS "code_postal",
         -- TODO: use position__city_code
         -- currently the field contains a majority of postal codes...
-        NULL                             AS "code_insee"
+        -- update(2024-08-07) : this is still the case.
+        NULL                                                     AS "code_insee"
     FROM lieux
     ORDER BY 1
 )