Configurably not destroy server, and use it next time

[OJFord]

While iterating to get something working, the spin-up time and 5m wait before shut down is quite frustrating.

To leave the server running and check some things manually, no amount of Ctrl-c is enough, it seems to require killing the top packer process.

It'd be good to be able to set it to stay up, and then on the next packer build identify it (perhaps by having set the label on Vultr) and reuse it rather than spin-up a new one.

[OJFord]

Packer does have -on-error=abort and -on-error=ask (default is cleanup) but the Vultr builder seems to ignore that and always destroy.