OS installations does not have init job machines selected

[skipi]

Describe the bug

On OS installation(CE/EE) - if you visit organization settings -> initialization jobs:

It looks as if the machine is already selected. This is not true - when you attempt to run a job or edit a workflow, you will get a 500 with logs like:

 ** (exit) an exception was raised:                                                                                                      
     ** (MatchError) no match of right hand side value: {:ok, {:error, :fetching_job_creation_failed}}                                   
         (front 0.0.1) lib/front_web/controllers/project_controller.ex:233: FrontWeb.ProjectController.render_default_branch/1           
         (stdlib 4.3.1.6) timer.erl:235: :timer.tc/1                                                                                     
         (watchman 0.2.0) lib/watchman.ex:36: Watchman.benchmark/2                                                                       
         (front 0.0.1) lib/front_web/controllers/project_controller.ex:3: FrontWeb.ProjectController.action/2                            
         (front 0.0.1) lib/front_web/controllers/project_controller.ex:3: FrontWeb.ProjectController.phoenix_controller_pipeline/2       
         (phoenix 1.6.16) lib/phoenix/router.ex:354: Phoenix.Router.__call__/2                                                           
         (front 0.0.1) lib/front_web/endpoint.ex:1: FrontWeb.Endpoint.plug_builder_call/2                                                
         (front 0.0.1) lib/front_web/endpoint.ex:1: FrontWeb.Endpoint."call (overridable 3)"/2

Steps to reproduce

1. Create a new project
2. Click edit workflow button
3. See 500 error page

Additional context

To fix this, one needs to click the Save changes button in the initialization jobs tab. It looks like the machine type is not selected at all, but it's appearing as selected only because there's no "Empty value" in the dropdown.

[TomFern]

I've faced this issue while setting up a CE instance for the Semaphore-sponsored devops course. From the end users perspective, it seems Semaphore is completely broken. There is no helpful message and right now it's not documented so people testing Semaphore are likely to give up at this point.

I've seen this issue on v1.2.3, v1.3.0 and v1.4.0 (on CE, I've not tested EE yet). I had to downgrade our course instance to v1.1.2, which works correctly.