Self-deployed server / client architecture with UI and built-in resource management

As per the [initial proposal](https://github.com/ConsenSys/storm/blob/master/Proposal.md):
- Deploys itself + Management UI and backend
- Executed only once
- `--production` is non-reversible / destroys local keys
- Management UI and backend
- Accounts for Spokes (and eventually external users)
- Dogfooding solves hosting the management infrastructure
- Resource/billing limits set globally, per account and/or spread by groups
- Launches clusters and containers
- Control from UI or CLI tool
- Cloud providers pre-configured from initial config
- Choose allowed types of instances
- Keep UI synced with cluster state
- Built-in monitoring with cAdvisor+InfluxDB+Grafana
- cAdvisor composed in each container
- InfluxDB on backend receives monitoring data
- Grafana dashboard to visualize container resource usage
- Alerting (Kapacitor? Probably needs a more complete solution)