Signed-off-by: Vitalii Koshura <[email protected]>

Instead, use a WSL container containing the docker engine,
and send it commands using the WSL mechanism

An image (especially in the universal model) may contain WU-specific files,
or maybe completely a function of the WU.

It's OK if we create lots of images.
In general they'll share most of their layers
so there's no disk penalty.

Remove container and image at end of job
User docker log to should output from container

In standalone mode, slot dir is . and project is project/

Add some verbose output; send all debugging output to stderr

We don't use Docker Desktop.
Instead we use a WSL Ubuntu 24.04 image in which podman is installed.

    Either or both could be present on both Linux and Win/WSL.
    If both, use podman.
    Do this separately for engine and compose;
    a host could have the podman engine but Docker compose.

data structures:
old: (in both HOSTINFO and WSL_DISTRO):
    bool is_docker_available: it's present, AND allowed by config
    string docker_version
    ... and same for docker_compose
new:
    string docker_version: empty if not present
    DOCKER_TYPE docker_version_type (DOCKER or PODMAN)
        We send this in sched requests even if Docker disallowed by prefs
        (send that flag separately).
    This way projects can know what software a host has,
    even if they can't use it.
    (Should use this approach with VBox too)

… requests.

(still query the presence of docker, and report it in sched requests).

scheduler: parse the above

docker_wrapper: use either docker or podman, depending on what's there
(on the host (Unix) or in a WSL distro (Win))

use pause/unpause instead of start/stop
Note: this works with podman only if root;
need to figure out how to do that

needed for podman's pause and unpause to work

Add a lib function run_command_pipe() for this purpose.