build_root_plan now returns None for non-linked (main) worktrees,
so archiving a thread that lives on a main worktree no longer
attempts `git worktree remove` (which git rightfully refuses).

The thread is simply marked archived in the metadata store and can
be unarchived later without any git operations.

Since build_root_plan now guarantees a linked worktree,
RootPlan::worktree_repo is changed from Option<Entity<Repository>>
to Entity<Repository>, removing the fallback path and the
now-unnecessary is_some() guard in archive_worktree_roots.

Cherry-pick of #53127 to stable

----
Self-Review Checklist:

- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX

checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [ ] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable

Closes #53056.

Release Notes:

- Fixed terminal tabs losing keyboard focus after switching tabs on
Linux X11

Co-authored-by: Andre Roelofs <p.arroelofs@gmail.com>

Remove the has_unbacked_linked_worktree filter that was too aggressively
hiding threads whose folder_paths included linked worktree paths not yet
discovered by git state. Since threads are already correctly grouped via
their main_worktree_paths column, this filter was unnecessary and caused
threads belonging to open workspace roots and closed worktree workspaces
to be hidden.

Update test_git_worktree_added_live_updates_sidebar to expect the thread
visible from the start (the chip name is derived from the path heuristic
even before git discovery).

Fix test_worktree_add_key_collision_removes_duplicate_workspace to switch
back to workspace A before the collision, since the collision code keeps
the active workspace and drops the incoming duplicate.

This PR updates the agent thread UI to disable the thread feedback
controls based on the organization's configuration.

Closes CLO-629.

Release Notes:

- N/A

When removing a project group, the fallback closure calls
find_or_create_local_workspace which searches self.workspaces for an
existing match. Because the workspaces to be removed haven't been
removed yet at that point, workspace_for_paths can return a doomed
workspace, hitting the assert in MultiWorkspace::remove.

Fix by adding an `excluding` parameter to find_or_create_local_workspace
so callers inside removal fallbacks can skip workspaces that are about
to be removed. Both remove_project_group and sidebar::archive_thread
now pass the appropriate exclusion set.

When a user on editor layout clicks "Try Now" on the parallel agents
announcement, the callback dispatches ToggleFocus in the same flush
cycle as set_layout(Agent). Because set_layout writes to the settings
file asynchronously, ToggleFocus opens the agent panel in its old dock
(right dock in editor layout). When the settings later reload, per-panel
migration observers fire in registration order. If the git panel
migrates to the right dock before the agent panel migrates out, the git
panel takes the active slot, and the agent panel's was_visible flag
becomes false when it subsequently moves to the left dock—leaving the
git panel visible and the agent panel hidden.

Fix: when switching layouts, defer the ToggleFocus dispatch until after
the settings change propagates by using a one-shot observe_global
observer on SettingsStore. Global observers fire in registration order,
so by the time this observer runs, all panel-migration observers
(registered at workspace creation) have already moved every panel to its
correct position.

Update the way `pane::RevealInProjectPanel` is handled to ensure that,
regardless of whether the file belongs to any open project, the project
panel is always activated and focused.

This refactor is a result of some internal feedback after changing its
handling so as to show a notification stating that the item that the
user was trying to reveal didn't belong to an open project.

We feel users are probably already used to relying on `cmd-shift-e` (on
macOS), in pretty much every context, in order to open the project
panel, and so having situations where it doesn't actually happen seems
like a bad user experience.