fetch messages

Downloads messages by UID and parses each into the same object mail parse produces : addresses, subject, text, html, attachments, headers. Does not mark them read unless mark seen is on.

Fetches the given uids from mailbox (default INBOX) and outputs messages: [{ uid, flags[], seen, answered, flagged, deleted, draft, internal_date (epoch ms), size, message, parse_error, raw? }] plus count, and message = the first entry for the single-uid case. message is the mail_parse object: { from {name,email}, to[], cc[], reply_to[], subject, date, message_id, in_reply_to, references[], text, html, attachments[{name,content_type,size,content_base64,inline,cid}], headers, size, spam_score, authentication }. A message the parser cannot read is still listed with message=null and parse_error set, so a flow that moves or deletes what it handled never skips a uid. A uid the mailbox no longer holds is simply absent. mark_seen=false (default) uses BODY.PEEK and leaves unread state untouched : a flow polling 'unseen' must set it, or mark the uids with imap set flags, or it will see the same messages every run. headers_only=true fetches the header block only (no bodies, no attachments). include_content=false drops attachment bytes. include_raw=true adds the raw RFC 822 text. Bounds: at most 500 uids per fetch, a message over max_message_bytes (default 25 MB, ceiling 50 MB) is never downloaded, and one fetch pulls at most 100 MB in total; a message refused by either bound is STILL listed, with skipped set and message null, so a flow can move or delete it by uid without reading it : never silently omitted. headers_only fetches are exempt from the byte bounds. Suppressed in a simulation.