Optimize getMembers query in member llm workflow

[skwowet]

Changes proposed ✍️

Simplified member data retrieval by replacing nested LEFT JOINs with a correlated subquery for organizations. This change significantly improves query performance for members with high activity counts (many identities/organizations).

Before:

-- Uses nested LEFT JOINs and derived table
LEFT JOIN "memberOrganizations" mo on mem.id = mo."memberId"
LEFT JOIN (select ...) as organizations on organizations."memberId" = mo."memberId"

After:

-- Uses correlated subquery
SELECT jsonb_agg(jsonb_build_object(...))
FROM "memberOrganizations" mo
JOIN organizations o ON o.id = mo."organizationId"
WHERE mo."memberId" = mem.id

Impact:

• 3-4x faster for members with high activity estimate
• Eliminates double joins (prevents unnecessary row duplication)
• No data loss - verified against production data