feat(#2075): migrate marts__combined_course_enrollment_detail to dimensional models

[quazi-h]

What are the relevant tickets?

Closes #2075 (mart migration — depends on prerequisite PR #2317)

⚠️ This PR must be merged and deployed AFTER #2317 (dim_course_run and tfact_enrollment column additions, deployed with --full-refresh). Deploying this mart before #2317 will result in NULL values for courserun_upgrade_deadline and courserunenrollment_is_edx_enrolled.

Description (What does it do?)

Migrates marts__combined_course_enrollment_detail from intermediate combined models (int__combined__courserun_enrollments, int__combined__course_runs, int__combined__users) to the dimensional layer.

The 6-way per-platform UNION ALL is replaced by a single flat query joining dimensional models:

• tfact_enrollment (filtered to enrollment_type = 'course') — enrollment base + enrollment_is_edx_enrolled (added in feat(#2075): add courserun_upgrade_deadline to dim_course_run and enrollment_is_edx_enrolled to tfact_enrollment #2317)
• dim_course_run (filtered to is_current = true) — course run attributes + courserun_upgrade_deadline (added in feat(#2075): add courserun_upgrade_deadline to dim_course_run and enrollment_is_edx_enrolled to tfact_enrollment #2317)
• dim_course — course_title and course_readable_id (via dim_course_run.course_fk)
• dim_user — user demographics and all platform-specific user IDs/usernames
• tfact_grade — grade data joined on user_fk + courserun_fk
• int__combined__courserun_certificates — certificate data (ecommerce order CTEs retained from int__ models)

Notable join strategy decisions:

• Certificate join: uses user_email (from dim_user) as the join key for all platforms. Bootcamps username is not stored in dim_user, so email is the reliable cross-table key.
• edX.org MITxOnline certificate cross-lookup: previously used int__micromasters__users CTE; now uses dim_user.user_mitxonline_username directly, eliminating that CTE.
• edX.org MicroMasters order join: uses dim_user.micromasters_user_id replacing the micromasters_users CTE user_id lookup.
• Bootcamps order join: changed from user_username-based to user_id-based (dim_user.bootcamps_application_user_id = bootcamps_completed_orders.order_purchaser_user_id), since Bootcamps username is not in dim_user.
• MITxPro orders: bridged via a supplemental mitxpro_enrollment_orders CTE sourcing ecommerce_order_id from int__mitxpro__courserunenrollments (not yet in tfact_enrollment).
• Grades: sourced from tfact_grade joined on user_fk + courserun_fk — clean dimensional join, no per-platform grade CTEs needed.

All 38 output columns preserved and verified via systematic diff against the original final SELECT.

How can this be tested?

Prerequisite: #2317 must be deployed first (with --full-refresh on both dimensional models).

-- Q1: Row count parity by platform
SELECT
    platform,
    count(*) AS row_count
FROM ol_warehouse_marts_qa.marts__combined_course_enrollment_detail
GROUP BY 1
ORDER BY 1;

-- Q2: Verify courserun_upgrade_deadline is populated for MITxOnline
SELECT
    platform,
    count(*) AS total_rows,
    count(courserun_upgrade_deadline) AS non_null_upgrade_deadline
FROM ol_warehouse_marts_qa.marts__combined_course_enrollment_detail
WHERE platform = 'MITx Online'
GROUP BY 1;

-- Q3: Verify courserunenrollment_is_edx_enrolled distribution
SELECT
    platform,
    courserunenrollment_is_edx_enrolled,
    count(*) AS row_count
FROM ol_warehouse_marts_qa.marts__combined_course_enrollment_detail
GROUP BY 1, 2
ORDER BY 1, 2;
-- Expected: edxorg and residential = all true; mitxonline/mitxpro = mix of true/false/null;
-- bootcamps/emeritus/global_alumni = all null

-- Q4: Verify grade data populates from tfact_grade
SELECT
    platform,
    count(*) AS total_enrollments,
    count(courserungrade_grade) AS enrollments_with_grade,
    count(courserungrade_is_passing) AS enrollments_with_pass_flag
FROM ol_warehouse_marts_qa.marts__combined_course_enrollment_detail
GROUP BY 1
ORDER BY 1;

Additional Context

Deployment order (no breaking change if followed):

1. Merge and deploy feat(#2075): add courserun_upgrade_deadline to dim_course_run and enrollment_is_edx_enrolled to tfact_enrollment #2317 — run dim_course_run --full-refresh and tfact_enrollment --full-refresh
2. Verify Q1/Q2 queries in feat(#2075): add courserun_upgrade_deadline to dim_course_run and enrollment_is_edx_enrolled to tfact_enrollment #2317 show correct values
3. Merge and deploy this PR — run marts__combined_course_enrollment_detail normally (no full-refresh needed)

Downstream consumers (dashboards, reports) using courserun_upgrade_deadline or courserunenrollment_is_edx_enrolled will see no regression as long as step 1 completes before step 3.