Implement configuration about parsec

[hug-dev]

The idea sounds good to me!

• What should be the final structure of the config file?

An alternative structure with indentation would be:

[providers]
  [providers."mbed-with-persistence"]
  type = "mbed-userspace"
  persist_keys = true
  key_id_manager = "on-disk-manager"
  converter = "protobuf"

  [providers."mbed-without-persistence"]
  type = "mbed-userspace"
  persist_keys = false
  key_id_manager = "on-disk-manager"
  converter = "protobuf"

The names mbed-wth-persistence are not needed for this file but could be logged somewhere and provide a way to add indentation on this file. I don't know which one is more readable.
In the listener section there should probably be a type as well in case different listeners are implemented?

• Should we move to generating provider ID dynamically?

I think it could be done in a later change with its specific design for simplicity.

• Do we offer the option of setting the provider ID in the configuration?

As the Key ID Manager uses the Provider ID in the path where it stores the Key ID mappings, a provider needs to have the same ID when the service restarts. For this reason, I guess it is a good idea to explicitely set an ID in the configuration.

• What happens if we have duplicate authenticators? (e.g. two "simple" authenticators with different default app names?)

Forbid having duplicate authenticators?

• Do we allow defaulting of parameters? (e.g the converter for providers would always be "protobuf", at least for a while)

I am in favor of that.

[hug-dev]

Aside note:

From man page daemon (7), it is said that systemd daemons can receive the SIGHUP signal to reload their configuration.
I am unsure of the security implications of that but for example, when this signal is received the threads could be joined, the current instances could be dropped and the configurator could be called again. That would be a nice feature 🍸

[hug-dev]

As per #50 it would be nice to add a logging option in the configuration that contains a binary option to have or not timestamps in the logs. I think by default it should be disabled.

[hug-dev]

Is there more work to do after #51 ?

[ionut-arm]

I think the work done so far should cover this issue. If we need to do more things we can raise new issues specific to what other configurables will be added.