[DPE-10875] refactor(config): inject the resource provider instead of routing via state

[marceloneppel]

Issue

CharmState carries a resource_provider slot that the charm assigns after construction, and an available_resources property that raises RuntimeError when nobody set it. ConfigManager is the only consumer, so the state class is acting as a service locator for a single manager's dependency. Raised by @akram09 in review on #182 and again on #183.

Solution

Hand the provider to ConfigManager directly and drop both the slot and the property, so the "not set" failure mode stops being representable. Each charm exposes it through a get_resource_provider() method (abstract on the base charm, returning the K8s manager or the VM workload), and the manager resolves the getter on use.

It is injected as a getter rather than an instance because the K8s manager that supplies it is constructed after ConfigManager in the charm's __init__ — the same late binding the existing charm bridges already use, so no construction reordering is needed. The injected type is the existing ResourceProvider protocol rather than a K8s client, since the VM charm satisfies it through its workload and K8sManager is not importable on VM.

Checklist

• I have added or updated any relevant documentation.
• I have cleaned any remaining cloud resources from my accounts.

[akram09]

Thank you !