Wait out SQLITE_BUSY when loading CloudSync on new pool connections.

[ComputelessComputer]

after_connect gave up on BEGIN IMMEDIATE after ~270ms, so sqlx retried the whole connect and logged "database is locked" during normal desktop writes.

---

Note

Medium Risk
Touches connection initialization and SQLite locking behavior on every new pool connection with CloudSync enabled; wrong retry/classification could delay connects or mask real load failures.

Overview
CloudSync sqlite-sync extension loading no longer wraps BEGIN IMMEDIATE around each load. That transaction fought active writers and caused sqlx after_connect to fail quickly (~270ms) with "database is locked" when the pool grew during normal writes.

Loads now use a short PRAGMA busy_timeout = 50 so SQLite’s C busy handler does not block Tokio workers, then retry load_extension + validation asynchronously with exponential backoff for up to 5 seconds (aligned with app-wide busy handling). Busy detection was extended via is_busy_error so extension-load failures that surface as locked-database messages are retried, not only raw sqlx busy codes.

The db-core integration test for concurrent pool acquires under a held writer was adjusted: it no longer depends on a fixed 90ms delay before commit, matching the new willingness to wait on busy during connect.

^{Reviewed by Cursor Bugbot for commit 08460eb. Bugbot is set up for automated code reviews on this repo. Configure here.}

[netlify]

✅ Deploy Preview for anarlog canceled.

Name	Link
🔨 Latest commit	413845e
🔍 Latest deploy log	https://app.netlify.com/projects/anarlog/deploys/6a8654845e5a7a00084291f4