Automation analysis export

Everything needed to analyze a flow in a single document: overview, metrics per node, DAG structure (the node ids the other tables refer to), clicked links with their role, enrolled contacts row by row with the node they are on, and the log of steps traversed. Each table is CSV, with its definition next to it: 'opens' without saying they are human ones, and 'clicks' without saying they exclude the logo, are numbers an analyst reads at random.

Without dataset you get a single text file with all tables; with dataset you get that raw CSV. Test runs are excluded everywhere. Tables with one row per contact are cut at 10,000 (enrollments) and 50,000 rows (steps): the cut is declared in the file name and in the document, never silently.

Email addresses appear only if the key also carries contacts:read: automations:read alone returns no personal data anywhere else, and an export must not be the shortcut around that rule. Without it, rows stay identified by contact_id. Required scope: automations:read.