The partial-failure test only placed the unknown ID after the valid
exit node, so it never verified that selection processing continues
past a leading invalid ID. Run the same assertions for both orderings.
Selecting all routes wipes every explicit selection, so exit nodes fall
back to management's auto-apply flags, which may mark several at once.
Reconcile immediately after select-all so at most one stays active
instead of waiting for the next network map.
A partial selection failure (e.g. an unknown ID in the request) still
selects the valid routes, so run the sibling exit-node deselection
regardless of the error and report both failures together.
Move route select/deselect handling from the daemon server into exported
routemanager methods (SelectRoutes, DeselectRoutes, SelectAllRoutes,
DeselectAllRoutes) so every consumer shares one implementation: v4/v6
exit-pair expansion, exit-node mutual exclusion, and selection triggering.
Previously the exit-node exclusivity lived only in the daemon's
SelectNetworks RPC, so the Android and iOS bindings could leave two exit
nodes selected until the next network map reconciliation. Both bindings
now call the shared manager methods and enforce exclusivity at toggle
time, matching the desktop behavior.
Zoltán Papp