### System details

* Omarchy 4.0.1-1 (Quattro)
* Hyprland 0.56.2-1
* NVIDIA RTX 3080 Ti
* Two DisplayPort outputs: 2560x1440 at 180 Hz and 200 Hz
* Linux 7.1.9

### What is wrong?

After suspending while the native Omarchy session lock is active, a restored or recreated lock surface can show a password field that does not own keyboard focus. Focus can also be lost while PAM disables the field and then releases it after a failed attempt.

This is separate from the slow-display reblank problem in #7399. It is also separate from the first modifier event lost during input-device restoration, which was later isolated in hyprwm/aquamarine#375.

### Steps to reproduce

1. Suspend through the normal Omarchy path so the session is locked before sleep.
2. Resume and wait for the displays to return.
3. Inspect or interact with the visible password field after output recreation.
4. The field can lack active focus. The same loss can occur after PAM rejects a password and re-enables the field.

### Root cause

`LockView.qml` requested `passwordInput.forceActiveFocus()` only when the component completed or `inputEnabled` changed. During resume, ext-session-lock outputs can be recreated while `inputEnabled` remains true. `Service.qml` already observes `Quickshell.onScreensChanged`, but did not ask the recreated view to reacquire keyboard focus.

On a multi-monitor resume, outputs can return sequentially, so one immediate refocus is not sufficient. PAM also temporarily disables the input during authentication, requiring recovery to pause and restart when the field becomes enabled again.

### Expected

The password field owns keyboard focus whenever the visible lock surface is ready for input, including after output recreation and after a failed PAM attempt.

### Verified fix

A cloned `omarchy.lock` service sends a focus-generation request when the lock becomes secure, whenever screens change, and after PAM releases the input. `LockView` is a `FocusScope`, acknowledges `TextInput.activeFocus`, and the service retries briefly across sequential output recreation. Focus telemetry confirmed that the field reacquired active focus.

The separate case-sensitive first-password failure was not used to validate this fix; protocol traces attributed it to aquamarine#375.

A focused regression test covers the retry budget, screen-change trigger, failed-PAM transition, disabled-field guard, focus scope, request handoff, and focus acknowledgement. PR #9181 contains the fix.