Helper functions for managing MLModels in Elasticsearch

Useful when you've trained your model on a development cluster and want to export and reimport into a production cluster.

Was thinking a `to_json()` which takes a `buf` and has an argument `compress` which gzip-compresses? Then we can have `MLModel.from_json()` as the import function.

cc @Winterflower @stevedodson for additional ideas here

The issue names no files, tests, or entry points; start by tracing how MLModel objects are currently represented and how Elasticsearch model import/export is handled. Clarify the serialization format, buffer and compression behavior, and the expected round-trip between development and production clusters before implementation; done means the agreed export and import workflow works reliably.

Written by the indexing model from the issue text.