Official plugins should support remote query/write on distributed clusters

Nobody has claimed this yet.

In a distributed cluster with separate processor, querier, and ingestor nodes, the plugins assume the processor node can also query and write locally.

For example, with the [Basic Transformation ](https://github.com/influxdata/influxdb3_plugins/tree/main/influxdata/basic_transformation)plugin, scheduled execution needs to both query source data and write transformed output. On a processor-only node, I hit these errors:

>Node mode without query cannot retrieve table chunks executing query

and

>cannot write to a read-only server: must enable ingest mode

This makes official plugins difficult to use in distributed deployments unless the processor node also has query and ingest enabled.

It would be helpful if official plugins supported remote query and remote write as an option, so a processor node could talk to a querier for reads and an ingestor for writes over HTTP.

For example, something like additional trigger arguments such as:

>query_mode=remote_http
>write_mode=remote_http

In my case, I already have the remote endpoints specified via environment variables so the plugin knows where to connect::
`INFLUXDB3_QUERY_HOST_URL`
`INFLUXDB3_WRITE_HOST_URL`

That would make the plugins work much better in a multi-node cluster without requiring process/query/ingest to be colocated.

Start with the Basic Transformation plugin in influxdata/basic_transformation and trace its scheduled query and write paths. Review how INFLUXDB3_QUERY_HOST_URL and INFLUXDB3_WRITE_HOST_URL are exposed to plugins, then determine how the proposed remote_http trigger arguments should behave. Done means a processor-only node can run the plugin with remote query and write endpoints.

Written by the indexing model from the issue text.