tf-ci-gateway: move execution from master to docker TF-A node

The execution of top level jobs, i.e. tf-gerrit-tforg-l1,
tf-gerrit-tforg-l2, etc., and tf-ci-gateway jobs may exhaustate the
amount of Jenkins master runners and deadlocks can occur. To avoid this
problem, run tf-ci-gateway under the TF-A docker
node (docker-amd64-tf-a-bionic) so less master runners are use when
top level jobs are triggered. Tested at staging [1]

[1] https://ci.staging.trustedfirmware.org/job/lsandov1-tf-ci-gateway/855/

Signed-off-by: Leonardo Sandoval <leonardo.sandoval@linaro.org>
Change-Id: I36fe390096f7f7529567e6896428150509907205
diff --git a/tf-ci-gateway.yaml b/tf-ci-gateway.yaml
index 9b77882..4f0ab35 100644
--- a/tf-ci-gateway.yaml
+++ b/tf-ci-gateway.yaml
@@ -13,7 +13,7 @@
             wipe-workspace: false
 - job:
     name: tf-ci-gateway
-    node: master
+    node: docker-amd64-tf-a-bionic
     project-type: freestyle
     concurrent: true
     disabled: false
@@ -78,7 +78,7 @@
     publishers:
         - postbuildscript:
             builders:
-                - role: MASTER
+                - role: SLAVE
                   build-on:
                       - SUCCESS
                       - FAILURE