Improve policy validation

[lgfa29]

This PR improves policy validation in a few ways:

1. validate the right target component: the initial implementation only validated ScalingPolicy.Target which is a map[string]string and doesn't actually need validation.
2. encapsulate the logic to validate an HCL block in a function: this allows us to re-use the validation logic in different places, such as strategy and target.
3. return error messages related to the jobspec when possible: previously the error messages would be related to the specific internal structure of the policy, which is not significant to end users. New error messages use the same block1->block2 notation from the Nomad docs.
4. only test validateScalingPolicy: this is the main function that calls all the others. Testing individual sub-functions failed to capture some cases.
5. expand test cases: test cases now have a full policy as input, as opposed to selectively remove fields from a known valid input. This makes it more clear to understand what is being tested.
6. show validation error message in tests: it's useful to see how validation messages will be displayed to users. You can use the -show-validation-error flag when running tests to seem them:

go test -v -run Test_validate ./policy/nomad/ -show-validation-error

[cgbaker]

good stuff 👍