feat: render inline article images as markdown

[alextuan1024]

Summary

• render Draft.js atomic image blocks in twitter article output as Markdown images
• preserve current article text rendering for non-media blocks
• add regression coverage for dict-style and list-style article entityMap payloads

What changed

twitter article already converts article Draft.js blocks into Markdown, but it currently skips atomic blocks entirely. Those blocks can contain inline image entities, so article exports lose embedded pictures.

This patch adds best-effort extraction for article image entities and renders them as:

![caption](original image url)

It now supports both observed response shapes:

• dict-style content_state.entityMap
• list-style content_state.entityMap entries of the form { key, value }

It also resolves image URLs both from inline entity data and from article-level media objects when the inline entity only carries a mediaId.

Example

Before:

Intro

Outro

After:

Intro

![A cat](https://pbs.twimg.com/media/cat.jpg)

Outro

Validation

• uv run --extra dev pytest -q tests/test_client.py tests/test_cli.py tests/test_serialization.py
• uv run --extra dev ruff check twitter_cli/parser.py tests/test_client.py

Test plan

Unit / regression:

• dict-style entityMap with inline original_url renders Markdown image
• list-style entityMap with mediaId resolved through article.media_entities renders Markdown image
• existing article CLI / serialization tests still pass

Manual smoke tests with real articles:

• twitter article 'https://x.com/odyseus0z/status/2030416758138634583' --markdown
  • verifies inline article images render in the exported Markdown
  • expected: 2 Markdown image lines in body
  • dumped markdown file: harness-engineering-is-cybernetics.md
• twitter article 'https://x.com/elvissun/status/2025920521871716562' --markdown
  • verifies inline article images render in the exported Markdown
  • expected: 3 Markdown image lines in body
  • dumped markdown file: openclaw-agent-swarm-full-setup.md

Notes

• this is intentionally scoped to inline article images only
• it keeps the existing twitter article data model unchanged and only improves Markdown extraction

[alextuan1024]

Dear @jackwener,
Lemme know if you skipped images on purpose.
In that case, feel free to close this PR.

[jackwener]

Thanks again for the contribution. We cherry-picked the article inline image work into a new branch and finished the integration there so we can keep the recent atomic markdown fix from #37 intact as well. The follow-up PR is #38, and we also kept co-author credit for your work. To avoid duplicate review threads, we’re going to close this PR and continue from #38.

[alextuan1024]

Sure. Sounds great to me!