Flux queries of influxdb tags (keys(), tag groups) should allow push-downs of filters including _field in the predicate

Nobody has claimed this yet.

__Proposal:__
Flux queries corresponding to influxql meta-queries (`show tag keys`, `show tag values`) should support filtering by `_field` (the field name) efficiently. There is a performance improvement for these filters implemented in 2.x OSS that we would like to be back ported to 1.x.

See the tests proposed in https://github.com/influxdata/influxdb/pull/21160/files for how this feature would look.

Note the implementation in that PR would need to change somewhat to be compatible with InfluxDB Enterprise. Ideally the influxql backend could be shared by both flux and influxql - this would require improvements to the influxql metaquery backend.

__Current behavior:__
Flux queries over influxdb tag keys and values that otherwise could be pushed down to the storage engine and made more efficient will instead use less-efficient strategies when the filter() contains `_field`.

__Desired behavior:__
We would like these queries to be maximally efficient

__Alternatives considered:__
N/A.

__Use case:__
Filtering tag keys and values by which field keys they show up with.

Start with the tests proposed in the linked pull request and trace the Flux queries over tag keys and values that include _field in filter(). Confirm how the InfluxQL meta-query backend handles these queries and what Enterprise compatibility requires; done means the filters are pushed down efficiently for both Flux and InfluxQL-compatible paths.

Written by the indexing model from the issue text.