Add prometheus serializer and use it in prometheus output

[danielnelson]

Add a prometheus serializer, with support for round tripping metric_version=2 of the prometheus input.

related #4414

Use this serializer in the Prometheus output when metric_version=2. Formally deprecate metric_version = 1 in both the input and the output plugin.

closes #4415

Add use_batch_format support to the file output, similar to support in the amqp and kafka outputs. This is a move away from the Serialize function towards the SerializeBatch function in the Serializer interface, which allows for more efficient writing and writing multiple metrics to a destination. In regards to the prometheus serializer, the batch format is required in order to group metric families and to write histogram and summary types correctly.

There is a known bug introduced in this pull request, since histogram or summary metrics are split over multiple telegraf.Metric's, it is possible that a batch of metrics will contain only part of a histogram/summary . I documented this in the serializer docs. This issue makes me wonder if we need a way to "group" metrics for the purposes of batching?

Required for all PRs:

• Signed CLA.
• Associated README.md updated.
• Has appropriate unit tests.

[reimda]

Is there a full telegraf->prometheus->telegraf test? Maybe I missed it.

[danielnelson]

I'll do a roundtrip test as a follow-up pr since it will involve some changes to testutil which will then require quite a number of unittest changes.