Make matrix-synapse-reverse-proxy-companion not report "502 Bad Gatew…

…ay" when Synapse workers restart

Since nginx 1.27.3, we can make use of the `resolve` parameter for an `upstream`'s `server`,
to allow DNS resolution to happen continuously at runtime, not just once during startup.

Previously, this was not possible to do in an `upstream` block without
an nginx-plus subscription. Outside of an `upstream` block, we've used
and still use `set $backend ..` workarounds to get DNS resolution at
runtime, but now we can do it in `upstream` as well.

roles/custom/matrix-synapse-reverse-proxy-companion/templates/nginx/conf.d/matrix-synapse-reverse-proxy-companion.conf.j2

@@ -16,10 +16,18 @@
 {% macro render_worker_upstream(name, workers, load_balance) %}
 {% if workers | length > 0 %}
 	upstream {{ name }} {
+		{#
+			We need to use a zone so that the upstream is stored in shared memory,
+			otherwise we can't use `resolve` below, as reported by nginx:
+			> resolving names at run time requires upstream ".." in ... to be in shared memory
+		#}
+		zone {{ name }} 64k;
+
 		{{ load_balance }}
 		keepalive {{ ((workers | length) * 2) | string }};
+		resolver {{ matrix_synapse_reverse_proxy_companion_http_level_resolver }} valid=5s;
 		{% for worker in workers %}
-			server "{{ worker.name }}:{{ worker.port }}";
+			server "{{ worker.name }}:{{ worker.port }}" resolve;
 		{% endfor %}
 	}
 {% endif %}