GPUs and cluster support

0.3.0

• Fixed broken passing of zone and project arguments in gce_vm, gce_vm_template, gce_get_external_ip, and gce_set_metadata.
• remove gce_auth() to favour auth with JSON key (#79)
• Fix project-id error if numbers in project (#72)
• Block users using "rstudio" as a login name
• Remove defunct example from gce_schedule_docker
• Support GPU images for Tensorflow, keras etc. (#101) via gce_vm_gpu() and gce_vm(template = "rstudio-gpu") (#101)
• Support common instance metadata by supply gce_set_metadata(instance = "project-wide")
• Support minCpuPlatform in instance creation and via gce_set_mincpuplatform() (#112)
• Add ability to specify a startup-script in gce_vm_container()
• Switch RStudio templates to use startup-scripts and metadata
• Switch to applying a nginx proxy service to deal with port routing for templates
• Add gce_startup_logs() to track whats going on when launching an instance
• Vectorise gce_vm_delete, gce_vm_stop, gce_vm_start and gce_vm_reset functions so you can pass in a list of instances
• Add gce_vm_cluster() to make it easier to create clusters for future