Enhance GPU probing and performance tuning in provisioning

- Introduced GPU probing functionality to gather and store GPU specifications in `/mnt/swarm_data/.gpu-rent-gpu.json`, aiding in performance tuning.
- Updated `install_ollama.sh` and `install_llamacpp.sh` to utilize GPU information for configuring optimal runtime parameters.
- Enhanced `provision.py` to include GPU probing and performance tuning logic, ensuring better resource allocation for LLM operations.
- Improved documentation in `decisions.md`, `llm.md`, and `swarmui.md` to reflect changes in GPU handling and performance tuning processes.
- Added new tests to validate the GPU probing and model resolution logic, ensuring robustness in handling various GPU configurations.

src/gpu_rent/session.py

@@ -21,7 +21,7 @@ from gpu_rent.cloud import (
     wait_volume,
 )
 from gpu_rent.bootstrap import run_bootstrap
-from gpu_rent.provision import provision_vm
+from gpu_rent.provision import provision_vm, tune_swarm_perf
 from gpu_rent.ready import wait_backend_idle
 from gpu_rent.snapshot import ensure_boot_snapshot
 from gpu_rent.notify import notify_ready
@@ -119,6 +119,13 @@ def _bind_access(
         wait_backend_idle(cfg, ip, log)
     except CloudError as exc:
         log(f"ready: {exc}")
+    # Comfy venv + Backends.fds exist after Idle — sage/triton + ExtraArgs.
+    try:
+        if tune_swarm_perf(cfg, ip, log):
+            log("systemctl restart swarmui (perf ExtraArgs)")
+            run_ssh(cfg, ip, "sudo -n systemctl restart swarmui", timeout=120)
+    except Exception as exc:
+        log(f"perf tune: {exc}")
     try:
         ensure_boot_snapshot(
             conn,