Refactor Nomad policy parsing and add tests

[lgfa29]

This PR makes the logic that parses a policy from Nomad into the internal policy representation the Autoscaler uses.

The new parser makes a few assumptions:

1. scaling->policy->evaluation_interval defaults to the value of scan_interval default_evaluation_time from the agent config file.
2. the final Policy.Target.Config map has the values from scaling->policy->target->config (which are defined in the jobspec) merged with the values from api.ScalingPolicy.Target (which are predefined by the Nomad API as the group, job and namespace that has the scaling stanza). Values in the jobspec have higher priority.
3. the values of 'Job' and 'Group' from api.ScalingPolicy.Target are copied over to Policy.Target.Config as job_id and group if they are present yet (i.e., if the jobspec has job_id the value won't be copied over).

Open questions:

1. evaluation_interval is used in the jobspec and scan_interval is used in the agent config. Should one of them be renamed so they match? scan_interval renamed to default_evaluation_time.
2. Renaming Job and Group seems unnecessary, but there are other parts that rely on them being job_id and group, so I didn't want to change them now. Using them as-is would require users to write them capitalize in the jobspec, which is unusual. Maybe have some automatic case conversion?

Closes #99
Closes #98
Closes #71

[cgbaker]

might put the default above in the initializer.

[cgbaker]

on my phone, can't trace this back, but wanted to make sure that this is only expected to be set if "config" existed

[lgfa29]

Yes, this moving values from targetMap["config"] (which is a map[string]interface{}) into configMapString (which is map[string]string).

So if "config" is not set we don't need to do anything.

[jrasell]

tested locally and this resolves #99 #98 #71

my only comment would be that the parsing error log formatting is a little weird:

2020-05-04T09:53:10.923+0200 [ERROR] policy_handler.policy_handler: invalid policy: 1 error occurred:
	* scaling->policy->source must be string, found []interface {}

: policy_id=f99ca263-7a1e-bc3a-b331-7db33053173f

not sure if there is a way to improve this, or its a by-product of multierror.

[lgfa29]

tested locally and this resolves #99 #98 #71

my only comment would be that the parsing error log formatting is a little weird:

2020-05-04T09:53:10.923+0200 [ERROR] policy_handler.policy_handler: invalid policy: 1 error occurred:
	* scaling->policy->source must be string, found []interface {}

: policy_id=f99ca263-7a1e-bc3a-b331-7db33053173f

not sure if there is a way to improve this, or its a by-product of multierror.

Yeah, that's not great. These are validation error messages, so I will improve them in a new PR.