OCPBUGS-62790: Resize /var tmpfs to 10GiB for ABI installations

Agent-based installations on vSphere with 16GB RAM were failing with
"no space left on device" errors during ostree image operations. The
live ISO environment uses a tmpfs mounted at /var that is sized at 50%
of available RAM. On systems with 16GB RAM, this provides only 8GB of
tmpfs space.

During the bootstrap process, node-image-pull.sh creates a temporary
ostree repository in /var/ostree-container/repo to pull and apply the
node image. This operation has a peak tmpfs usage of approximately
8.5-9GB, exceeding the available 8GB and causing ENOSPC errors.

This fix resizes the /var tmpfs to 10GiB before creating the temporary
ostree repository, providing sufficient space for the image operations
while maintaining compatibility with the minimum 16GB RAM requirement.

The resize is performed using systemd-run to escape the mount namespace
and only affects systems running in the live environment (detected by
the presence of /run/ostree-live).

Assisted-by: Claude Code

Author: zaneb

data/data/bootstrap/files/usr/local/bin/node-image-pull.sh.template

@@ -34,6 +34,11 @@ if test -f /run/ostree-live; then
     mkdir -p "${ostree_repo}"
     echo "In live environment; creating temporary repo to pull node image"
     ostree init --mode=bare --repo="${ostree_repo}"
+
+    echo "Resizing /run/ephemeral_base tmpfs to 10GiB for ostree operations..."
+    # Use systemd-run to avoid inheriting MountFlags
+    systemd-run --wait --service-type=oneshot mount -o remount,size=10G /run/ephemeral_base
+
     # if there are layers, import all the content in the system repo for
     # layer-level deduping
     if [ -d /ostree/repo/refs/heads/ostree/container ]; then