Adds retry/exponential backoff logic and fixes Glide build

[ssalevan]

Thanks so much for writing this plugin, we've been using it to publish container metrics for our cloud-native platform, Spine.

This PR adds several changes to adapt to recent Prometheus Pushgateway changes:

• Removes timestamps from submitted metrics (the Pushgateway now rejects these)
• Adds configurable retry, timeout, and exponential backoff logic to improve robustness
• Adds configurable Prometheus job and instance identifiers, defaults to a job of 'unused' and the identifier automatically determined by the Pushgateway
• Fixes the infamous logrus Glide conflict bug

I've tested this change against our fleet and all unit tests passed cleanly.

Let me know what you think and thanks much!