fix(message): render markdown while streaming assistant output

[SuMuxi66]

Summary

Fix assistant messages to render Markdown while streaming, instead of showing raw Markdown text until the assistant turn completes.

Before

During streaming, assistant output was rendered as raw text, so users could temporarily see Markdown syntax such as:

• | table |
• **bold**
• code spans

Markdown formatting only appeared after the turn finished.

After

Assistant output always renders through the Markdown component.
When the message is still streaming, the cursor is appended after the rendered Markdown.

Files changed

• desktop/frontend/src/components/Message.tsx

Test checklist

• Streaming tables render as Markdown instead of raw |
• Bold / code spans render during streaming
• Streaming cursor still appears while the answer is in progress
• Cursor disappears after the turn completes
• Final Markdown output still renders correctly
• Copy button behavior is unchanged
• Reasoning fold/unfold still works

Scope

This PR only changes assistant message rendering while streaming.
It does not change the Composer, past:chats, file references, or the CLI/TUI flow.

[SivanCola]

Thanks, this is the right direction and the scope is much cleaner than the broader streaming-markdown work in #3140. I think this PR should be the main path for the focused assistant streaming markdown fix.

Before merging, please fix the streaming cursor placement. Markdown renders a block-level .md wrapper, so rendering <span className="cursor" /> as a sibling after <Markdown /> places the cursor after the markdown block rather than at the end of the rendered text. For normal paragraphs this will show the cursor on the next line, not after the last token.

Please make the cursor render as part of the final rendered markdown content, then verify paragraphs, lists, tables, and code blocks while streaming.

Once this is addressed and verification stays green, I will consider merging it into main-v2.