Inventory Enhancements and Reorganisation

[jacderida]

• 2cbf1de chore: remove terraform check on inventory cmd

  It shouldn't be necessary to have an initialised Terraform workspace to get the inventory for
  deployment, which is the case when using a GHA workflow.

• 819abb5 refactor: narrow binary option for deployment

  We previously had an SnCodebaseType enum with three variants: Main, Branch and Versioned.
  Now that we can use the release repository to retrieve the latest versions for binaries, I didn't
  see the necessity for a third option; we simply use a version or we build from source. With that in
  mind, I renamed the enum to BinaryOption and provided BuildFromSource and Versioned variants.

  I don't think there should be any functional change here.

• 08e1d4f refactor: introduce inventory module

  I was finding the main lib module difficult to navigate, so I decided to move inventory-related
  code into its own module, and introduced a new DeploymentInventoryService struct for encapsulating
  the generation of inventory.

  I also made changes and additions in the ansible module, introducing concepts from our own domain
  rather than making it completely generic. We now pass enums for playbooks and inventory values, so
  we don't have to assemble paths all over the place. I think this makes the code more compact,
  readable and maintainable.

• 7604a20 chore: remove --safe-version arg from deploy

  We don't need to specify the client version when we are using a versioned deployment. It only gets
  used if we are running automated smoke tests, which we don't do at the moment anyway. In any case, I
  updated the smoke-test and upload-test-data commands to supply a --safe-version argument, in
  case we do end up using these again in the future.

  Also changed the use of the Error type in the manage_test_data module in favour of just using
  eyre. It's debatable that this whole crate should be changed to do this and remove the lib
  module.

• 8720c3f chore: use the node registry to populate inventory

  Now that we have access to the sn_service_management crate, we don't need small structs to
  represent a subset of data from the node registry. We can deserialize the fetched node registries
  directly to a NodeRegistry struct.

  Some other modifications:

  • Fix some Clippy warnings.
  • Rename codebase_type variables to binary_option.
  • Tidy up the sections in the printed report.
  • The BinaryOption is made optional when generating the inventory, because we only have access to
    it on the machine where the deployment initially ran. If it is not supplied, we just use the
    version numbers from the genesis node. We would lose the initial branch information if the testnet
    was deployed using a branch specification, but in most cases that should be for experimental
    testnets.
  • Remove node_count in favour of peers.len().
  • For some reason safenodemand_endpoints was storing a peer ID and socket address pair, but the
    daemon doesn't have a peer ID, so I just removed this.
  • The peer multiaddresses are in the NodeRegistry, so we don't need to fetch these via SSH as an
    additional step.

• 8f6fc45 chore: review feedback

  • Change inventory_type argument to use a reference to AnsibleInventoryType to avoid cloning in
    loops.
  • Centralise check for the existence of inventory files.
  • Remove irrelevant documentation