[DT-4263][2/7] Add an email sends endpoint for the Admin Console email log

[kevinmarete]

Addresses

https://broadworkbench.atlassian.net/browse/DT-4263

Security risk: low — a new admin-only read endpoint; it adds recipients' display names to data /api/mail/range already returns to admins.

Summary

Each recipient of an email gets its own email_entity row, so the Email Log shows a digest sent to 87 people as 87 rows. GET /api/mail/sends returns one entry per send instead, with its recipient count and names, for the grouped Email Log agreed on DT-4084.

• Stacked on [DT-4263][1/7] Record a send id on each email row #3149, which records a send_id per request. A send is the rows sharing a send_id, email type and entity reference. Rows written before [DT-4263][1/7] Record a send id on each email row #3149 have no send_id, so for them a send is the rows of one type and reference each created no more than 10 minutes after the previous one.
• Same start, end, limit (capped at 1000) and offset as /api/mail/summary, paged by send, newest first. Each send lists up to 100 recipients by name, each marked sent or not (SendGrid accepted it; nothing records inbox delivery); recipientCount counts them all.
• /api/mail/summary is unchanged: the deployed Email Log page still reads it, until part 3 moves the page to /sends.

Reading the query

• in_range: the rows in the date range. lone_id is set only on an older row with neither a send_id nor an entity reference, so it partitions alone.
• flagged: per type, reference, lone_id and send_id, starts_send = 1 on the first row. Rows with a send_id never start another; older rows do after a gap of more than 10 minutes.
• numbered: a running sum of starts_send, so every row of one send shares a send_number.
• grouped: each row gets send_row_id, the smallest email id in its send, which identifies the send from here on.
• page: one row per send with its time and distinct recipient count, then this request's OFFSET/LIMIT.
• named: the page's rows joined to users, one row per recipient, ordered by display name, sent if SendGrid accepted any of their copies.
• The final SELECT builds each send's recipients from its first 100 names.

Example: two NEW_STUDY_DIGEST rows from one daily job share a send_id 30 minutes apart, so they're one send. Two NEW_DAR rows for DAR-1 from different requests a minute apart have different send_ids, so they're two.

The default end is now the start of tomorrow in the server's time zone, the same zone explicit dates are read in; this also applies to /range and /summary, and changes nothing where the server runs in UTC.

The 10-minute cutoff hasn't been measured against dev data yet. A send that crosses the range's start or end is split there.

---

Have you read CONTRIBUTING.md lately? If not, do that first.

• Label PR with a Jira ticket number and include a link to the ticket
• Label PR with a security risk modifier [no, low, medium, high]
• PR describes scope of changes
• Get a minimum of one thumbs worth of review, preferably two if enough team members are available
• Get PO sign-off for all non-trivial UI or workflow changes
• Verify all tests go green
• Test this change deployed correctly and works on dev environment after deployment

🤖 Generated with Claude Code

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
100.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud