fix(week10): correct WARN-severity guidance and add INNER JOIN hint

Two student-facing fixes found by building the assignment against live
Azure Postgres:

- Singular-test stub claimed severity is "set in dbt_project.yml". The
  template ships no dbt_project.yml (students dbt init their own), and a
  project-level severity would silently downgrade the not_null /
  unique_combination PK tests. Point students at an inline
  {{ config(severity='warn') }} instead.
- Mart stub said only "join trips to zones". A pickup_location_id with no
  matching zone (e.g. 999) becomes NULL borough under a LEFT JOIN and fails
  the not_null PK test. Tell students to INNER JOIN so pickup_borough stays
  non-null and the mart lands at the rubric's 184 rows.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PLwBsooLgWEkUqJYVK7FsU

Author: Lasse Benninga

models/marts/fct_daily_borough_stats.sql

@@ -14,7 +14,10 @@ zones AS (
 )
 
 SELECT
-    -- TODO: join trips to zones on pickup_location_id = location_id
+    -- TODO: join trips to zones on pickup_location_id = location_id.
+    --   Use an INNER JOIN: a few trips have a pickup_location_id with no matching
+    --   zone (e.g. 999). INNER JOIN drops those so pickup_borough is never NULL and
+    --   can serve as part of the mart's primary key (your not_null test needs this).
     -- TODO: aggregate to grain (pickup_borough, pickup_date)
     -- Required output columns:
     --   pickup_borough   TEXT     - z.borough

tests/assert_avg_tip_pct_within_bounds.sql

@@ -3,11 +3,15 @@
 -- which is unusual and almost always indicates a small-sample bucket (e.g. the
 -- Unknown borough) where a few high-tip outliers dominate the average.
 --
--- This test is intentionally set to WARN severity in dbt_project.yml so it
--- does not block the build, but it surfaces the anomaly for the reports/answers.md
--- write-up. The assignment rubric requires documenting this finding.
+-- Set this test to WARN severity by adding an inline config at the top of this
+-- file (below these comments):  {{ config(severity='warn') }}
+-- That keeps a few expected Unknown-borough rows from blocking `dbt build`, while
+-- still surfacing them for your reports/answers.md write-up (the rubric requires
+-- documenting this finding). Do NOT set a project-level test severity in
+-- dbt_project.yml: that would also downgrade your not_null and
+-- unique_combination primary-key tests, which you want to stay at ERROR.
 --
--- The test passes when zero rows are returned; any returned rows are flagged.
+-- The test passes (no WARN) when zero rows are returned; any returned rows are flagged.
 
 -- TODO: write the SELECT here.
 -- Query {{ ref('fct_daily_borough_stats') }} and return rows where avg_tip_pct > 1.