[DOCS] Adds sync to data frame transform API

[lcawl]

This PR adds the sync property to the create data frame transform API reference (https://www.elastic.co/guide/en/elasticsearch/reference/master/put-data-frame-transform.html)

[elasticmachine]

Pinging @elastic/es-docs

[elasticmachine]

Pinging @elastic/ml-core

[hendrikmuhs]

there is another level between sync and field/delay, its "sync" -> "time" -> "field"/"delay" (see your example below)

sync defines continuous mode, time defines that the sync method is using a timestamp field in the data. In future we will have other ways to sync source and dest, e.g. for non timeseries data

[lcawl]

Thanks, I've drafted a description of that time property.

[hendrikmuhs]

we identified yesterday that we need a default value for delay, I will update this comment once we agreed on a value.

[sophiec20]

Adding prev discussion notes for visibility

Please highlight that we recommend that you select a field that represents time of ingest for your time field. This is because we use the field to check which entities have changed since the last time we checked. If you have an ingest time field, then we can easily find what has changed. If you use a different time field e.g. time of event, then you may need to make sure delay is set to a large enough value to allow for any delays due to the elapsed time it takes for data to reach the system... This is a link to an example for using an ingest pipeline to set an ingest timestamp https://www.elastic.co/guide/en/elasticsearch/reference/7.2/accessing-data-in-pipelines.html#accessing-ingest-metadata

The delay default is 60s.

[lcawl]

Thanks for the feedback @hendrikmuhs and @sophiec20

[sophiec20]

LGTM