add nginx keepalive_timeout env to prevent race condition on gorouter

[loadpi]

the current default of 30 causes the same inconsistent 502's as https://community.pivotal.io/s/article/5004y00001buMQz1626802995951?language=en_US

and we need to set it to >90s so the gorouter is in charge of closing the connections
https://docs.cloudfoundry.org/adminguide/routing-keepalive.html

tl;dr

Reason for this change:

With GoRouter's 90s keep alive enabled and the build pack's 30s it caused a intermittent EOF errors on the gorouter because it was trying to use an already closed connection...

{"log_level":3,"timestamp":"2023-05-24T09:26:55.408725231Z","message":"backend-endpoint-failed","source":"vcap.gorouter","data":{"route-endpoint":{"ApplicationId":"zzz","Addr":"zzz","Tags":{"app_id":"zzz","app_name":"xastest","component":"route-emitter","instance_id":"0","organization_id":"zzz","organization_name":"zzz","process_id":"zzz","process_instance_id":"zzz","process_type":"web","source_id":"zzz","space_id":"zzz","space_name":"zzz"},"RouteServiceUrl":""},"error":"EOF","attempt":1,"vcap_request_id":"zzz","retriable":false,"num-endpoints":1,"got-connection":true,"wrote-headers":true,"conn-reused":true,"dns-lookup-time":0,"dial-time":0,"tls-handshake-time":0}}

and on the client this header

'x-cf-routererror': 'endpoint_failure (EOF)'

Here are the two links as ref:

• Keepalive Considerations
  https://docs.cloudfoundry.org/adminguide/routing-keepalive.html

• Similar behaviour on spring apps
  https://community.pivotal.io/s/article/5004y00001buMQz1626802995951?language=en_US

Testing this change:

Without any changes:

CF-nonp-4> cf push --no-start -b https://github.com/loadpi/cf-mendix-buildpack.git\#Add-nginx-keepalive-timeout -m 2G -k 4G -p SimpleAppRetrieves.mda xastest -d dev.mendixcloud.com
CF-nonp-4> cf start xastest
CF-nonp-4> cf ssh xastest
$ cat app/nginx/conf/nginx.conf | grep _timeout
    keepalive_timeout 30;

Then with the envvar

CF-nonp-4> cf set-env xastest NGINX_KEEPALIVE_TIMEOUT "100"
CF-nonp-4> cf restage xastest
CF-nonp-4> cf ssh xastest
$ cat app/nginx/conf/nginx.conf | grep _timeout
    keepalive_timeout 100;

This app starts posting to /xas twice every 10s and on long running tests with puppeteer the error's are not present when the nginx keepalive timeout is 100s