fix: streaming Export leaves a recoverable .partial file on cancel/error

[BorisTyshkevich]

What & why

Closes #105.

Cancelling an export (or hitting a mid-stream server error) called writable.abort() on the File System Access API's FileSystemWritableFileStream. On Chrome this leaves a hidden, 0-byte .crswap swap file behind and never materializes the visible target the user picked — so a cancelled/failed export recovered neither the target file nor any partial data, just an invisible orphan.

streamToFile (src/ui/app.js) now close()s the writable in the catch path instead — finalizing whatever bytes were already committed under the target handle — then renames it in place to <name>.partial via FileSystemFileHandle.move() (Chrome 110+), so a cancelled/failed export leaves a clearly-labeled, inspectable partial artifact. Falls back to leaving the plain, non-renamed file when .move() is unsupported, or if the rename itself throws (e.g. a name collision) — best-effort, never blocks surfacing the original error.

Applies uniformly to both export paths that share streamToFile: the single-file export (exportDirect) and the per-statement script export (exportScript).

Checklist

• npm test passes (the per-file coverage gate is non-negotiable)
• Tests added/updated in the same change as the code
• npm run build succeeds (single-file dist/sql.html)
• Layers kept honest: pure logic in src/core/, network in src/net/ (injected fetch), DOM in src/ui/
• No new runtime dependency
• README / CHANGELOG.md ([Unreleased]) updated if behavior or the deployed surface changed
• Reconciled affected tracked work (roadmap Roadmap to 1.0.0 #68, the issue body, ADR/CHANGELOG) if this change reshaped it — n/a beyond the CHANGELOG entry; this is a standalone inbox fix, not on the Roadmap to 1.0.0 #68 build order.

Test plan

• npm test — 1281/1281 passing, all per-file coverage gates hold (app.js: 100/96.24/92.01/100 — within its documented floor)
• npm run build — dist/sql.html builds clean
• Two independent review passes (correctness + test-quality) found no bugs; two test-quality nits (duplicated fixture, missing abort-not-called assertions) were applied
• Not run through Playwright e2e — no existing e2e spec touches Export (native file pickers aren't exercised there), and this change doesn't touch any rendered UI, only an already-seamed error path already covered by unit tests with mocked FileSystemWritableFileStream/FileSystemFileHandle

🤖 Generated with Claude Code