[agent] refactoring the shell agent

[feckert]

The agent is at first sight a bit confusing.
I found it difficult to read in there. How everything works.
So I would like to make some changes to improve the maintainability.
I have already staged a pullrequest for refactoring the init script. The refactoring of agent is a little more difficult. So I would like to share my plans first.

• The following lines should go into a new function to parse the command line arguments https://github.com/openwisp/openwisp-config/blob/master/openwisp-config/files/openwisp.agent#L4-L37

• The following lines should go into a new function called main. This is the agent it self. This is more or less C style. https://github.com/openwisp/openwisp-config/blob/master/openwisp-config/files/openwisp.agent#L608-L658 The new main function is called at the end of the script with main "$@"

• All log messages are encapsulated in a new function. This makes the code clearer.

• If the functions have correct names (which is most the case) then unnecessary comments can be removed. This also makes the code clearer.

• And I would like to check the code with the tool shellcheck and fix the warnings.

• The ubus call returns json so we should use in the following lines the json wrapper from openwrt to parse the output.
  https://github.com/openwisp/openwisp-config/blob/master/openwisp-config/files/openwisp.agent#L115-L117

Those would be my first thoughts. I don't think that's the last.

I'm asking for feedback on what you guys think about that.