Allow distributing facet instances into multiple subnets in VPC

[garo]

Inside an EC2 VPC the target availability zone is controlled by specifying the subnet instead of directly specifying an availability zone.

The current VPC implementation is both fragile as it requires that the availability_zones matches the subnet and it doesn't allow the facet instances to be distributed into multiple subnets.

I'm proposing the following changes:

• When vpc is set then availability_zones must not be set.
• We change subnet into subnets which is an Array of all subnets where the instances are distributed.

Then when launching a facet with multiple servers then the facet instances are distributed in a round robin fashion into all described subnets.

Also the computer list would display subnet instead of AZ for each server.