Refactor remaining roles to follow standard pattern

Applied standard role pattern to all remaining roles:

**artefact-manager**:
- Added state management (artefact_manager_state)
- Split tasks into main.yml (dispatcher), deploy.yml, undeploy.yml
- Converted kubectl commands to kubernetes.core.k8s module
- Added kubeconfig variable with fallback support
- Organized defaults with clear section headers

**homer**:
- Added state management (homer_state)
- Split tasks into main.yml (dispatcher), deploy.yml, undeploy.yml
- Converted shell commands to kubernetes.core.k8s module
- Added kubeconfig variable with fallback support
- Organized defaults with clear section headers

**zot**:
- Added state management (zot_state)
- Updated main.yml to route to undeploy.yml when absent
- Created undeploy.yml for cleanup tasks
- Converted kubectl namespace creation to kubernetes.core.k8s
- Replaced kind_config_dir with zot_kubeconfig throughout
- Organized defaults with clear section headers

**prometheus**:
- Added undeploy.yml with Helm uninstall and CRD cleanup
- Updated main.yml to route to undeploy.yml when absent
- Replaced kind_config_dir/kubeconfig_output_dir with prometheus_kubeconfig
- Converted kubectl namespace creation to kubernetes.core.k8s
- Organized defaults with clear section headers

**node-feature-discovery**:
- Added undeploy.yml for NFD removal
- Updated main.yml to route to undeploy.yml when absent
- Replaced kind_config_dir with nfd_kubeconfig throughout
- Converted kubectl namespace creation to kubernetes.core.k8s
- Organized defaults with clear section headers

All roles now:
- Support state-based deployment (present/absent)
- Use kubernetes.core.k8s instead of kubectl where possible
- Have consistent kubeconfig variable pattern with fallback
- Follow the standard structure defined in role-template/
- Pass ansible-lint with 0 failures and 0 warnings