Let devs choose the handler for http requests

[julien-nc]

So they can choose a stream handler for example.

Summary

When using the 'stream' => true request option with IClient, the response body is a resource but its content is written all at once when the request is done.
This happens because the use of the CurlHandler is hardcoded in https://github.com/nextcloud/server/blob/fb67d5d/lib/private/Http/Client/ClientService.php#L44
The Curlhandler makes a blocking call to curl_exec which executes the request synchronously.

To allow progressive streamed responses, we need:

• to use HTTP 1/*
• to allow guzzle to use the StreamHandler

This is just a PoC. I have no idea if it's the right thing to do.
Wdyt?

Checklist

• Code is properly formatted
• Sign-off message is added to all commits
• Tests (unit, integration, api and/or acceptance) are included
• Screenshots before/after for front-end changes
• Documentation (manuals or wiki) has been updated or is not required
• Backports requested where applicable (ex: critical bugfixes)
• Labels added where applicable (ex: bug/enhancement, 3. to review, feature component)
• Milestone added for target branch/version (ex: 32.x for stable32)

AI (if applicable)

• The content of this PR was partly or fully generated using AI

[provokateurin]

Have you tried if it works?

[julien-nc]

@provokateurin Yes, it is used in a WIP branch https://github.com/nextcloud/integration_openai/blob/enh/noid/streaming/lib/TaskProcessing/TextToTextProvider.php#L131-L137 . The output of the task processing task is progressively echoed to stdout.

There also are standalone scripts to test that. With the enh/noid/streaming branch of integration_openai, you can run:

• php -S 127.0.0.1:8088 apps-extra/integration_openai/tests/stream_i_client_sse_server.php that spawns a simple HTTP server returning a SSE stream.
• php apps-extra/integration_openai/tests/stream_i_client_repro.php http://127.0.0.1:8088/chat/completions that specifically tests the HTTP client's behaviour with 'stream' => true

You can try running those on server's main and on this PR's branch to see the difference. Spoiler, with this branch, the client progressively writes in the resource.