[josephleekl]

Deploys the Backline QPM from the openQSE/QFw Backline QPM PR on a new qec-backline-head node and exposes it to Slurm jobs as --qpu=backline-local.

• docker-compose.yml, slurm.conf, tools/provision-qfw-cluster.sh: new qec-backline-head slurmd node in the qfw-services partition.
• config/site-services.yaml: backline-local service definition.
• config/qfw-slurm/plugin.conf, config/qfw-slurm/resources.lua: map --qpu=backline-local to service backline-local.
• config/device-access.yaml, config/qpu-users.json: device record, and entitlement for the test users.
• modulefiles/backline, tools/install-backline-venv.sh: shared venv with Catalyst and PennyLane, plus a clone of qri-backline in /workspace/qfw-container-base/qri-backline.
• tools/qfw-site-services: --target backline start, stop and status. It's opt-in: all doesn't start it.
• docs/backline.md: setup, including the QFw development override from the backline-qpm branch, and how to submit the demo job (sbatch -A …/examples/job.sh).
• Tested: the demo job ran end to end on the local cluster. The executor stopped and the reservation was released at the end of the job.
• Known issue (pre-existing, not from this PR): current qfw-slurm main rejects the stock plugin.conf, because it maps two AWS resources to one service.

Merge order and external references

• Merge the openQSE/QFw Backline QPM PR first. docs/backline.md builds the QFw override from openQSE/QFw main, and the cluster clones openQSE/QFw-SLURM-Cluster main, so the setup only works once both are merged. Until then, use the backline-qpm branches of josephleekl/QFw and josephleekl/QFw-SLURM-Cluster.
• tools/install-backline-venv.sh clones the demo from josephleekl/qri-backline (personal repo; there is no openQSE copy). Change the URL if the demo moves under openQSE (QRI_BACKLINE_DIR already overrides the checkout path).

[josephleekl]

Moved to openQSE#64