Support EmbeddedAnsible SCM credentials

[Fryguy]

This is an attempt (along with ManageIQ/manageiq-schema#394 ManageIQ/manageiq-schema#395) to deal with the "double ownership problem" with ConfigurationScriptSource#authentication and GitRepository. The idea is that in the model we can delegate authentication stuff to the GitRepository. Thus, we ensure a git_repository exists, then sync the authentication and other git attributes over to it.

In GitRepository model, while it might have it's own authentication via AuthenticationMixin, we also add a separate belongs_to :authentication which is a referenced authentication. This way we can support both directly owned authentications and not-owned authentications.

cc @gtanzillo @carbonin @NickLaMuro @agrare

[Fryguy]

Note that this PR doesn't work as is, since authentication_id is coming in during create!, and it's failing since the git_repository doesn't exist yet.

[Fryguy]

My original plan was to just have the caller set the git_repository.authentication directly, but then never save it, to make it "temporary", however if anything subsequently calls a git_repository method, then the authentication is lost and that feels wrong.

[carbonin]

Where and how will we handle if the SCM credential changes for a configuration script source?
Should we create an after_save hook or something to also change the one on git repositories?

[carbonin]

Just caught this re: #19027 (comment)

[Fryguy]

This PR effectively makes scm_url a required attribute for a ManageIQ::Providers::EmbeddedAnsible::AutomationManager::ConfigurationScriptSource, which I think is ok. Without this line you will consistently get a validation error on a GitRepository#url format.

[Fryguy]

This PR is ready-to-go codewise, but I want to add some tests around the configuration_script_source -> git_repository interactions. I did a lot of testing manually in rails console, and want to reflect those in real specs.

[Fryguy]

Ok specs added. I had to wrap the existing spec in a context because it always created a rugged repo, but these new tests didn't need it. So, view this with whitespace diff hidden.

@carbonin Please review.

[Fryguy]

FYI I really tried to DRY this up with respect to the attrs_for_sync_git_repository, by using something like changed.include_any?, but because one side has :url and the other side has :scm_url it got really ugly, so I'd prefer to just leave it this way for now.

[Fryguy]

@carbonin This is ready to go.

[miq-bot]

Some comments on commit Fryguy@0deca62

spec/models/manageiq/providers/embedded_ansible/automation_manager/configuration_script_source_spec.rb

• ⚠️ - 143 - Detected allow_any_instance_of. This RSpec method is highly discouraged, please only use when absolutely necessary.
• ⚠️ - 170 - Detected allow_any_instance_of. This RSpec method is highly discouraged, please only use when absolutely necessary.

[miq-bot]

Checked commit Fryguy@0deca62 with ruby 2.4.6, rubocop 0.69.0, haml-lint 0.20.0, and yamllint 1.10.0
4 files checked, 3 offenses detected

app/models/manageiq/providers/embedded_ansible/automation_manager/configuration_script_source.rb

• ❗ - Line 53, Col 3 - Style/AccessModifierDeclarations - private should not be inlined in method definitions.
• ❗ - Line 65, Col 3 - Style/AccessModifierDeclarations - private should not be inlined in method definitions.
• ❗ - Line 72, Col 3 - Style/AccessModifierDeclarations - private should not be inlined in method definitions.

[simaishi]

Ivanchuk backport details:

$ git log -1
commit 0afb961d335f0bcb15d457c2d0d13029d604017b
Author: Nick Carboni <ncarboni@redhat.com>
Date:   Tue Aug 6 11:03:17 2019 -0400

    Merge pull request #19027 from Fryguy/embedded_ansible_scm_credentials
    
    Support EmbeddedAnsible SCM credentials
    
    (cherry picked from commit c26295952742657e6f8056ea6ba81508060095e0)