FR: inputs.http could support streaming with chunked encoding

### Use Case

I currently try to stream metrics in influx line protocol from one telegraf to the next. The receiving telegraf opens the connection to a python-script-webserver (outputs.execd) with bearer token authentication.

This all works well and good, but I have to batch the data and the requesting telegraf is tied to an interval in which there might be more or less metrics and there isn't a mechanism to speed up or slow down the interval dynamically.
So I just thought, why not stream the metrics?

But `[[inputs.http]]` insists on doing a gather-run per interval and I don't see how to switch it to streaming like `[[inputs.execd]]` can do.

So... Feature Request: Streaming with `[[inputs.http]]`, or maybe you have an alternative for me?

### Expected behavior

If the connection doesn't close (but in a sound way, like chunked streaming) then don't complain about timeouts but read the streamed data.

### Actual behavior

data streaming not possible.

### Additional info

_No response_

Start by locating the inputs.http implementation and its gather, timeout, and response-reading paths. Trace how it handles connections that remain open, then determine the boundaries of supporting chunked streaming without treating the connection as a timeout. Done means streamed metrics can be read continuously while preserving the expected inputs.http behavior.

Written by the indexing model from the issue text.