set lower "kubectl delete --grace-period" for dev cleanup

[ahmetb]

skaffold v0.8.0

This is semi-related to #770. I think when someone is doing dev, it's most likely OK to do a kubectl delete --grace-period=1 seconds (which I believe overrides pod's terminationGracePeriodSeconds and if that's unspecified, kubernetes default of 30 second for this value).

This would provide faster cleanups (and therefore exits) if/when cleanup stage properly waits for all resources to go away (as pointed out in #770).