Render vm_memory_high_watermark for RabbitMQ

RabbitMQ/Erlang do not properly determine available RAM when running in
containerised environment, plus default value for watermark is 0.4, which is
low for nodes running RabbitMQ only.

Some additional information:
https://www.rabbitmq.com/memory.html

Change-Id: I1738c66be0e698568881b2c64d5d456db851af4d
Depends-On: Iadd0529d166ef1dad1697215bbc181104a7b926e
Depends-On: Ibc6e8b91055f411b058505d14c2d83fbc826cdd4

service/files/defaults.yaml

@@ -18,6 +18,7 @@ configs:
     # How many seconds we give a node before making liveness checks
     # completely strict
     node_startup_timeout: 180
+    high_watermark_ratio: 0.75
 
 url:
   rabbitmq:

service/files/rabbitmq.config.j2

@@ -23,6 +23,7 @@
                     {verify,verify_peer},
                     {fail_if_no_peer_cert,false}]}
      {% endif %}
+     ,{vm_memory_high_watermark, {absolute, {{ (memory_limit|int * rabbitmq.high_watermark_ratio)|int }}}}
    ]}
   ,{autocluster, [
       {dummy_param_without_comma, true}