mlx: refined model push behavior (#15431)

* mlx: refined model push behavior

Refine the algorithm for parallel push of safetensors based models to get
better reliability and throughput.

* review comments, hardening, and performance tuning for slow links

* review comments

cmd/cmd.go

@@ -2009,7 +2009,7 @@ func appendEnvDocs(cmd *cobra.Command, envs []envconfig.EnvVar) {
 Environment Variables:
 `
 	for _, e := range envs {
-		envUsage += fmt.Sprintf("      %-24s   %s\n", e.Name, e.Description)
+		envUsage += fmt.Sprintf("      %-27s   %s\n", e.Name, e.Description)
 	}
 
 	cmd.SetUsageTemplate(cmd.UsageTemplate() + envUsage)
@@ -2473,6 +2473,7 @@ func NewCLI() *cobra.Command {
 				envVars["OLLAMA_CONTEXT_LENGTH"],
 				envVars["OLLAMA_KEEP_ALIVE"],
 				envVars["OLLAMA_MAX_LOADED_MODELS"],
+				envVars["OLLAMA_MAX_TRANSFER_STREAMS"],
 				envVars["OLLAMA_MAX_QUEUE"],
 				envVars["OLLAMA_MODELS"],
 				envVars["OLLAMA_NUM_PARALLEL"],