fix: handle escaped pipe characters in markdown table cells

[Br1an67]

TLDR

Fixes table rendering when cells contain literal | characters (e.g. A|B). Previously, the naive .split("|") incorrectly broke such content into separate columns. Now uses a regex split on unescaped pipes with \| → | unescaping.

Dive Deeper

Root cause: MarkdownDisplay.tsx lines 114 and 130 used .split("|") to parse table rows, which cannot distinguish between structural column delimiters and literal pipe characters in cell content.

Fix: Added a splitTableRow() helper that:

1. Splits on unescaped pipes using negative lookbehind: (?<!\\)\|
2. Unescapes \| → | in the resulting cell text
3. Trims whitespace from each cell

LLMs can output A\|B in table cells and it renders correctly as A|B in a single column. This follows the standard Markdown table escaping convention.

Reviewer Test Plan

1. Ask the model to output a table where cells contain | (e.g. union types like string|number)
2. In the system prompt or QWEN.md, instruct the model to escape pipes as \| in tables
3. Verify the table renders with correct column alignment

Testing Matrix

	🍏	🪟	🐧
npm run	❓	❓	❓
npx	❓	❓	❓

Linked issues / bugs

Fixes #2461

[Br1an67]

Closing in favor of #2503 which has the cleaner implementation with a dedicated splitTableCells() helper.