fix(healthchecks) run tcp active healthchecks in stream module

Instead of disabling active healthchecks in the stream module
altogether (which was done to avoid duplicate active healthchecks),
what we need to do is to run the "tcp" type healthchecks in the
stream module, and the other ones ("http", "https") in the http
module.

kong/runloop/balancer.lua

@@ -300,12 +300,9 @@ do
 
       -- Do not run active healthchecks in `stream` module
       local checks = upstream.healthchecks
-      if ngx.config.subsystem == "stream"
-         and (checks.active.healthy.interval ~= 0
-              or checks.active.unhealthy.interval ~= 0)
+      if (ngx.config.subsystem == "stream" and checks.active.type ~= "tcp")
+         or (ngx.config.subsystem == "http" and checks.active.type == "tcp")
       then
-        log(ngx.INFO, "[healthchecks] disabling active healthchecks in ",
-                      "stream module")
         checks = pl_tablex.deepcopy(checks)
         checks.active.healthy.interval = 0
         checks.active.unhealthy.interval = 0