New scheduler configuration does not work well with non-standard server config

[bartv]

The scheduler uses the server.server-address to connect to the orchestrator server. This is documented as: The public ip address of the server. This is required for example to inject the inmanta agent in virtual machines at boot time. This setting is actually no longer useful because we no longer start agents remotely

However, in setup where you do not set bind-address to the public ip or you use the external ip of the loadbalancer the scheduler will not be able to connect and it does not clearly say this in log messages.

The setup that caused this was the following:

• cloudflared on the same host
• bind-address set to 127.0.0.1
• server-address to the public hostname orchestrator.example.com

The only hint in the logs was the agent and the server logs complaining about connection timeout. This was solved by setting server-address to 127.0.0.1

proposal

1. connect to the bind-adress, except if it is 0.0.0.0, then we connect to 127.0.0.1
2. make sure the error reporting is clear when the scheduler is not starting (e.g. refer to the correct log file from the server log)

[wouterdb]

Hey team! Please add your planning poker estimate with Zenhub @arnaudsjs @Hugo-Inmanta @jptrindade @sanderr