feat(server): keep monitors and in-flight runs going when a licence lapses

Monitor execution and the workflow runner are deliberately unguarded: billing
state must not take away a customer's ability to know their infrastructure is
on fire, and killing a run midway leaves a half-configured server.

The plan called for a server-limit check in gRPC Register. Left out: a server
row only comes from CreateServer, which already checks the cap, so counting in
Register counts the caller itself and would reject a legitimate agent at
exactly the cap.

server/internal/grpc/server.go

@@ -26,6 +26,16 @@ type vantageServer struct {
 	pb.UnimplementedVantageServer
 }
 
+// Register carries no licence check, deliberately.
+//
+// A server row only ever comes from CreateServer, which checks the cap before
+// issuing a pre-registration token. By the time an agent calls Register its row
+// already exists, so counting here would count the caller itself: an instance
+// sitting exactly at its cap would reject the very agent it just authorised, and
+// every re-registration after a reinstall would fail too.
+//
+// The cap is enforced where rows are created, which is the only place it can be
+// enforced correctly.
 func (s *vantageServer) Register(ctx context.Context, req *pb.RegisterRequest) (*pb.RegisterResponse, error) {
 	agentToken, err := services.RegisterServer(req.ServerId, req.PreRegToken, req.Hostname, req.IpAddress, req.OsInfo)
 	if err != nil {