feat(twitter): resolve reposts, expand t.co links and resume tweets pagination

[UncertaintyDeterminesYou4ndMe]

Description

Three additive capabilities for opencli twitter tweets, motivated by backfilling a user's timeline (thousands of reposts) into a local Markdown archive:

1. retweeted_tweet column — a repost row only carries the truncated "RT @user: …" string, the reposter as author, and no media. The original post is now surfaced from legacy.retweeted_status_result in the same shape as the existing quoted_tweet (id, author, name, full text, counts, created_at, url, media, and its own quoted_tweet). Tombstoned / unavailable originals yield null; is_retweet is unchanged.
2. --expand-urls (default false) — replaces https://t.co/… in text, quoted_tweet.text and retweeted_tweet.text with expanded_url, collected from legacy.entities.urls, the long-form note_tweet … entity_set.urls (links in long posts are only listed there) and the media entities (the trailing t.co of attached photos/videos).
3. --cursor / --cursor-file — resume pagination from a saved bottom cursor and persist the next one (empty file = timeline exhausted). Together with the existing --page-delay this makes gentle, resumable backfills possible without re-fetching pages or tripping the 429 limiter. Details:
   • with --cursor-file, --limit becomes a soft target: whole pages are returned so no tweet falls between two runs (cursors point at page boundaries);
   • if the first page of a run fails (e.g. HTTP 429) the resume point is written before the error is raised, so the caller can retry the same page after cooldown;
   • a resumed run that finds no tail returns [] instead of EmptyResultError.

Output is unchanged when the new flags are off, apart from the appended retweeted_tweet column. collectTweetUrlMap, expandTweetUrls and extractRetweetedTweet live in shared.js so timeline / search / list-tweets can adopt them later.

opencli twitter tweets @jack --limit 500 --expand-urls true --page-delay 4 --cursor-file ./jack.cursor -f json
# later: continue from where the previous run stopped
opencli twitter tweets @jack --limit 500 --cursor "$(cat ./jack.cursor)" --cursor-file ./jack.cursor -f json

Related issue: —

Type of Change

• ✨ New feature

Checklist

• I ran the checks relevant to this PR — npx vitest run --project adapter clis/twitter/ (560 passed) and the default gate npm test; the only failures are 6 pre-existing ones in src/cli.test.ts (browser tab targeting) that fail identically on an untouched main checkout in my environment.
• I updated tests or docs if needed — new cases in tweets.test.js (expand-urls, retweeted_tweet, tombstone, cursor resume + cursor-file, rate-limited resume, empty resumed tail) and shared.test.js (helpers); usage example added to docs/adapters/browser/twitter.md; cli-manifest.json rebuilt.
• I included output or screenshots when useful

Documentation

• Used positional args for the command's primary subject unless a named flag is clearly better
• Normalized expected adapter failures to CliError subclasses instead of raw Error

Screenshots / Output

Live run against my own logged-in account (--limit 5 --expand-urls true --cursor-file ./cur.txt -f json), one row abridged:

{
  "id": "2100136163877208198",
  "author": "snwkng50462474",
  "is_retweet": true,
  "text": "RT @AdrianPunk115: 本文分为上下两册，直接作为做网页动效的指南。 https://x.com/i/article/2098669987930210304",
  "retweeted_tweet": {
    "id": "2100136...",
    "author": "AdrianPunk115",
    "text": "本文分为上下两册，直接作为做网页动效的指南。 https://x.com/i/article/2098669987930210304",
    "url": "https://x.com/AdrianPunk115/status/...",
    "has_media": false
  }
}

./cur.txt afterwards holds the bottom cursor of the fetched page; a second run with --cursor "$(cat cur.txt)" continued from it, and reaching the end of the timeline left the file empty.