iOS: fold waitForFile into connectWithRetry for parity with Android

[gmaclennan]

Context

PR #52 fixes an Android-specific FileObserver wd-collision bug by dropping waitForFile entirely and folding the file-existence check into the connect retry loop. iOS does not have this bug — it already uses 50 ms polling rather than a DispatchSource-based observer (per agents.md:255) — but it still keeps waitForFile and connectWithRetry as two separate primitives.

Proposed change

Apply the same architectural cleanup on iOS:

• Remove waitForFile(atPath:timeoutSeconds:) from ios/NodeJSIPC.swift.
• Update connectWithRetry to retry on the connect-side IOError/POSIXError for both "file missing" (ENOENT) and "file exists but server not yet accepting" (ECONNREFUSED), bounded by a 30 s deadline.
• Delete ios/Tests/WatchForFileTests.swift. The existing NodeJSIPCTests cover the same end-to-end behavior (connect-after-delayed-bind path is what waitsForSocketFileCreation already exercises on Android, and the iOS suite likely has the equivalent — verify and add if missing).
• Update doc references in agents.md, docs/ARCHITECTURE.md, README.md once the iOS-side mentions become stale too.

Why this matters

• Consistency: both NodeJSIPC implementations share the same shape. Easier to reason about, easier to keep in sync when one side gets a bug fix.
• Less code: ~−100 lines (mirrors the Android side's −436 with proportionally smaller iOS test files).
• One race-prone primitive instead of two.

Why this is not urgent

iOS does not have the wd-collision bug. Polling 50 ms file-existence + retry-connect works correctly today. This is a refactor for consistency, not a bug fix.

References

• PR android: fold waitForFile into connect retry loop #52 (Android fix): android: fold waitForFile into connect retry loop #52
• Code review of android: fold waitForFile into connect retry loop #52 noted iOS parity as a sensible follow-up, not in scope for the bug fix.