Use mDNS hostname as launchd label and CLI connection name

[nirs]

Starting and using a VM required two different names:

launchctl start local.docker
ssh root@docker.local

The launchd Label was reverse-DNS (local.$VM_NAME) while the VM's mDNS name is $VM_NAME.local. launchd.plist(5) only recommends naming the plist file <Label>.plist; the Label itself is an independent identifier. Keep the plist at ~/Library/LaunchAgents/local.$VM_NAME.plist and set the Label to the mDNS hostname.

Use the same name for the Docker context and podman system connection, so start, SSH, and the container CLI all share one identifier:

launchctl start docker.local
ssh root@docker.local
docker context use docker.local

In the podman benchmarks, iterate the full names instead of stitching prefixes and suffixes in the loop body. Prefix CONTAINERS_MACHINE_PROVIDER on the commands that need it instead of exporting it into the shell.

After dnf update on Fedora 44, podman -c fails with:

Error: unable to connect to Podman socket: Get "http://d/v6.1.0/libpod/_ping":
ssh: rejected: connect failed (open failed)

Fedora selinux-policy 44.6 dropped the permissive type for sshd_session_t. SSH stream-local forwarding to /run/podman/podman.sock is denied:

sshd_session_t → var_run_t : sock_file { write }
sshd_session_t → container_runtime_t : unix_stream_socket { connectto }

The podman tutorial now shows that failure and installs a local CIL module over SSH with those two allows. The module is stored in the guest policy store, so it survives reboot. It is not in cloud-init.