Export the latest run's results as CSV

One text/csv attachment with a fixed column set per contact: phone_e164, name, external_reference, language, time_zone, status, disposition, attempt_count, next_attempt_at, callback_at, skip_reason, last_call_id, last_call_status, termination_cause, termination_detail, updated_at, then the last call's business outcome: workflow_status, workflow_outcome_type (review_candidate, informational, terminal_note), workflow_outcome (the outcome step the run landed on), workflow_outcome_label, workflow_interrupt (wrong_party, cease_contact, dispute, hardship, callback, human_request, language_change), extraction_status (empty when no extraction was requested; pending, running, ready or failed), extraction_schema_version (the schema version the cells were extracted under; the column layout follows the agent's current schema), recording_available and transcript_available (true/false), and one extraction.<field> column per field of the agent's post-call extraction schema (advisory model output; empty unless extraction_status is ready). In list order, bounded by the contact-list cap (50,000 rows); X-Skysay-Row-Count carries the row count and X-Skysay-Truncated says whether rows beyond the cap were left out. Uploaded variables are not echoed back. Free-text cells that start with a formula character are prefixed; phone_e164 is written verbatim (import that column as text in a spreadsheet). Bound to the credential's own organization: an organization_id query value is ignored and a credential without an organization is refused.