Fix empty dhparam.pem

[seroron]

When launch the container, the default dhparam.pem is cleared to 0 by shell redirection while creating a new dhparam.pem.

# docker exec -it 0841481db300 bash
root@0841481db300:/app# cd /etc/nginx/dhparam/
root@0841481db300:/etc/nginx/dhparam# ls -lha
total 0
drwxr-xr-x 1 root root 25 Dec 16 13:40 .
drwxr-xr-x 1 root root 35 Dec 16 12:18 ..
-rw-r--r-- 1 root root  0 Dec 16 13:40 dhparam.pem

So, nginx sometimes fails to initialize.

# docker logs 0841481db300
WARNING: /etc/nginx/dhparam/dhparam.pem was not found. A pre-generated dhparam.pem will be used for now while a new one
is being generated in the background.  Once the new dhparam.pem is in place, nginx will be reloaded.
forego     | starting dockergen.1 on port 5000
forego     | starting nginx.1 on port 5100
dockergen.1 | 2018/12/16 13:40:02 Generated '/etc/nginx/conf.d/default.conf' from 7 containers
dockergen.1 | 2018/12/16 13:40:02 Running 'nginx -s reload'
dockergen.1 | 2018/12/16 13:40:02 Error running notify command: nginx -s reload, exit status 1
dockergen.1 | 2018/12/16 13:40:02 Watching docker events
dockergen.1 | 2018/12/16 13:40:03 Contents of /etc/nginx/conf.d/default.conf did not change. Skipping notification 'nginx -s reload'

[kamermans]

Good catch @seroron - this looks good to me @jwilder

[arneschreuder]

Can this PR be merged. Should solve #1226

[kamermans]

FYI: I have spent an hour or two trying to come up with a test for this (ex: using Python pyinotify with Threaded/AsyncNotifier to watch the file), but there is no clean way to write a regression test for this in my opinion.

My feeling is that this PR is implemented properly and it should be merged ASAP to avoid more problems. It seems something changed recently that causes this bug to become more prominent, either in nginx or in docker, since it didn't start appearing until recently, and now even the Travis-CI tests are failing because of it.