mlxrunner: bound MLX loads by system free memory while other models are loaded

On Apple silicon the scheduler's free-memory figure for the GPU is the
Metal working set minus what Ollama's own runners report. It does not see
memory held by other applications, so a second MLX model can pass the fit
check on a machine that is already short of memory, and the load pushes
the system into swap and compression.

While other models are loaded, the MLX fit check now also bounds the
available memory by the system's free memory on shared-memory GPUs, the
same rule llama-server loads already apply. A miss evicts an idle model
and retries instead of starting the load. First loads are unchanged: with
nothing else loaded, the model loads against the working-set figure alone,
as both engines do today. The check also does not cover memory that grows
after load, such as KV caches and prefix-cache snapshots.

x/mlxrunner/client.go

@@ -301,11 +301,14 @@ func (c *Client) HasExited() bool {
 }
 
 // Load checks whether the model fits in GPU memory and starts the subprocess.
-func (c *Client) Load(ctx context.Context, _ ml.SystemInfo, gpus []ml.DeviceInfo, requireFull bool) ([]ml.DeviceID, error) {
+func (c *Client) Load(ctx context.Context, systemInfo ml.SystemInfo, gpus []ml.DeviceInfo, requireFull bool) ([]ml.DeviceID, error) {
 	if len(gpus) > 0 {
 		modelSize := c.memory.Load()
 		// We currently only use the first GPU with MLX
 		available := gpus[0].FreeMemory
+		if requireFull && gpus[0].Integrated && systemInfo.FreeMemory > 0 && systemInfo.FreeMemory < available {
+			available = systemInfo.FreeMemory
+		}
 		overhead := gpus[0].MinimumMemory() + envconfig.GpuOverhead()
 		if available > overhead {
 			available -= overhead