Improved Runner.mt:update_status() method

Previously, the "principle of single responsibility" was not observed in
the `update_status()` - the method should do one thing. The part related
to output has been moved to a higher level so that the principle is
respected.

This will allow you to work with internal test content (e.g servers)
before output to stdout.

Part of tarantool#304

luatest/runner.lua

@@ -328,6 +328,7 @@ function Runner.mt:start_test(test)
     self.output:start_test(test)
 end
 
+-- luacheck: push no unused
 function Runner.mt:update_status(node, err)
     -- "err" is expected to be a table / result from protected_call()
     if err.status == 'success' then
@@ -341,8 +342,8 @@ function Runner.mt:update_status(node, err)
     else
         error('No such status: ' .. pp.tostring(err.status))
     end
-    self.output:update_status(node)
 end
+-- luacheck: pop
 
 function Runner.mt:end_test(node)
     node.duration = clock.time() - node.start_time
@@ -465,6 +466,7 @@ function Runner.mt:invoke_test_function(test)
             end
         end
     end
+    self.output:update_status(test)
 end
 
 function Runner.mt:find_test(groups, name)