Write SQL for multi-account metrics

A consumer app allows one user to own multiple accounts. Use SQL to answer the following questions. Assume the database has these tables: `accounts` - `account_id` BIGINT — primary key - `user_id` BIGINT — the owner of the account - `created_at` TIMESTAMP — when the account was created - `last_visit_at` TIMESTAMP — the most recent time the account was visited `notifications` - `notification_id` BIGINT — primary key - `account_id` BIGINT — foreign key to `accounts.account_id` - `sent_at` TIMESTAMP — when the notification was sent - `read_at` TIMESTAMP — when the notification was read; `NULL` means the notification is still unread Assumptions: - One user can have many accounts. - One account can have many notifications. - All timestamps are in UTC. - A "multi-account user" is a user with at least 2 distinct accounts. - A user "has unread notifications" if at least one of their accounts has at least one notification with `read_at IS NULL`. Write SQL for both tasks below: 1. Count how many users have exactly 2 accounts, exactly 3 accounts, and 4 or more accounts. - Required output columns: `account_bucket`, `user_count` - Use bucket values: `'2'`, `'3'`, `'4+'` 2. Among multi-account users only, compute the percentage of users who currently have at least one unread notification across any of their accounts. - Required output columns: `users_with_unread`, `multi_account_users`, `pct_users_with_unread` - Return the percentage as a decimal between 0 and 1.