feature(ApplicationManager): handle podman connections

[axel7083]

What does this PR do?

This PR makes the application manager requires a ContainerProviderConnection argument when pulling a recipe. This connection will be propagated to the RecipeManager and the BuilderManager, to respect the connection to use.

This allow us to target a specific connection if multiple exists. For now we keep the existing behaviour of taking the first running.

Screenshot / video of UI

N/A

What issues does this PR fix or reference?

Fixes #1527
Requires for #1462

How to test this PR?

• unit tests has been provided

Manuallyy

• Start a recipe, stop, restart
• assert no problem

[axel7083]

Short demo

This PR does not expose UI, however using the dev tools, we can make some experiments

On my Linux fedora 40, I have a podman machine (qemu), a native and a remote (unix) connection.

Inside AI Lab I can open the dev tools and check those connections

Let's store them inside a connections variable

Let's start a code generation recipe on the qemu machine

studioClient.requestPullApplication({
    'modelId': 'hf.ibm-granite.granite-8b-code-instruct',
    'recipeId': 'codegeneration',
    connection: connections[2]
});

And now, let's start a text_to_audio recipe on the native connection

studioClient.requestPullApplication({
    'modelId': 'hf.ggerganov.whisper.cpp',
    'recipeId': 'audio_to_text',
    connection: connections[1]
});

Finally we have two pod running on two different machines

Let's look at the console

$: podman --connection=podman-machine-default pod ls
POD ID        NAME                              STATUS      CREATED        INFRA ID      # OF CONTAINERS
9634e7b0fabc  pod-whispercpp-app-1723476984746  Running     8 minutes ago  ce88673f9681  3
$: podman pod ls
POD ID        NAME                           STATUS      CREATED        INFRA ID      # OF CONTAINERS
066ece98d4de  pod-codegen-app-1723477339738  Running     2 minutes ago  fb0e0fac6ab4  3

[jeffmaury]

LGTM