Review findings:

1. This change makes the first file-open path slower, not faster. ChatInterface now lazy-loads CodeEditor, but the actual file read still starts from CodeEditor's useEffect after that module mounts. That turns the flow into: click file -> wait for editor module -> start readFile/getFileContentBlob. For the reported issue (opening files from chat is very slow), this is a regression in the critical path rather than an improvement.

2. The new close button only cancels in-flight file reads after CodeEditor has already loaded. The Suspense fallback can hide the preview UI, but it cannot cancel the lazy module load itself. So if the visible delay is dominated by editor initialization, this PR still does not let the user close the real loading work; it only closes the shell around it.

Build output also shows Vite warning that CodeEditor is still statically imported elsewhere, so the dynamic import here does not currently split it into a separate chunk. That makes the performance benefit even less likely.

Validation run locally:

• npm run typecheck
• npm run build

Updated this PR after rebasing onto the latest main and addressing the review concern about the chat file-open path.

What changed:

• switched chat file clicks from opening the heavy inline CodeEditor directly to opening the lightweight preview modal first
• kept Open in editor wired to the existing editor sidebar flow for full editing
• passed AbortController.signal through ChatContextFilePreview file/blob requests so closing the preview during load now cancels the in-flight request
• kept the existing abortable loading behavior in CodeEditor
• removed the chat-specific lazy-editor loading path from the critical open flow

Validation:

• npm run typecheck
• npm run build

Notes:

• the branch was rebased onto the latest origin/main and force-pushed
• local uncommitted package-lock.json changes were intentionally kept out of the PR