feat(ai-chat): render Markdown while assistant replies stream

[byteoxo]

Summary

AI Chat now renders Markdown while the assistant reply is still streaming, instead of showing raw text until the turn finishes. Existing code-block actions (Copy / Insert) and the rest of the chat UI stay the same.

What's included

• Stream assistant text through MarkdownView for the full reply, not only after isStreaming clears
• Keep unclosed fenced code blocks ( /~~~`) as open code blocks while tokens arrive, then mark them closed when the fence completes
• During streaming, render code with a lightweight monospaced view so Copy / Insert stay available without rebuilding SourceEditor on every chunk
• After the stream ends, switch code blocks to the existing syntax-highlighted editor
• Throttle re-parses (~48ms) while streaming and flush immediately when the stream completes
• Unit tests for open/closed fences, partial tables, headers/lists, and fence length matching
• CHANGELOG entry under [Unreleased]

Testing

• Unit: TableProTests/MarkdownBlockParserTests
• Unit: TableProTests/AIChatCodeBlockDetectionTests
• Manual: ask the model for a full Markdown sample (headings, bold/italic, lists, blockquotes, tables, thematic breaks, fenced code) and confirm progressive rendering while tokens stream
• Manual: open SQL/JS code blocks mid-stream still show as code blocks before the closing fence
• Manual: Copy and Insert on streaming and completed code blocks still work
• Manual: stream completion does not leave raw Markdown or a broken code-block chrome