Adjust TrafficServer timeouts to work for longer response times.

We had only been testing shorter timeouts in our test suite, but it
turns out to handle longer timeouts, we need to increase
the other TrafficServer timeouts to basically match the nginx read
timeout (otherwise TrafficServer was still timing out after the connect
timeout and retrying the request multiple times).

Author: GUI

src/api-umbrella/cli/read_config.lua

@@ -332,9 +332,9 @@ local function set_computed_config()
   -- router is what should trigger the initial timeout. This also prevents the
   -- proxies further back in the stack from thinking the client unexpectedly
   -- hung up on the request.
-  config["trafficserver"]["_connect_attempts_timeout"] = config["nginx"]["proxy_connect_timeout"] + 1
+  config["trafficserver"]["_connect_attempts_timeout"] = config["nginx"]["proxy_read_timeout"] + 1
   config["trafficserver"]["_transaction_no_activity_timeout_out"] = config["nginx"]["proxy_read_timeout"] + 1
-  config["trafficserver"]["_transaction_no_activity_timeout_in"] = config["nginx"]["proxy_send_timeout"] + 1
+  config["trafficserver"]["_transaction_no_activity_timeout_in"] = config["nginx"]["proxy_read_timeout"] + 1
   config["nginx"]["_initial_proxy_connect_timeout"] = config["nginx"]["proxy_connect_timeout"] + 2
   config["nginx"]["_initial_proxy_read_timeout"] = config["nginx"]["proxy_read_timeout"] + 2
   config["nginx"]["_initial_proxy_send_timeout"] = config["nginx"]["proxy_send_timeout"] + 2