Add benchmarks for new terms_enum API

[markharwood]

In 7.14 we introduced a new API to elasticsearch which will be used by Kibana to offer auto-complete of partially typed terms.

We'd like to keep an eye on the performance of this as the speed of this API is of the utmost importance - trying to offer word suggestions before the user types the next character on the keyboard.

The suggestion was that the HTTP logs dataset would be a representative dataset to track this.

Relates to elastic/elasticsearch#71550