fix: Update SFTConfig parameter

- Change max_seq_length to max_length in SFTConfig constructor
- TRL deprecated max_seq_length in Feb 2024 and removed it in v0.20.0
- Reference: huggingface/trl#2895

This resolves the SFT training failure in CI tests

Author: Nehanth

llama_stack/providers/inline/post_training/huggingface/recipes/finetune_single_device.py

@@ -469,7 +469,7 @@ def setup_training_args(
             use_cpu=True if device.type == "cpu" and not torch.backends.mps.is_available() else False,
             save_strategy=save_strategy,
             report_to="none",
-            max_seq_length=provider_config.max_seq_length,
+            max_length=provider_config.max_seq_length,
             gradient_accumulation_steps=config.gradient_accumulation_steps,
             gradient_checkpointing=provider_config.gradient_checkpointing,
             learning_rate=lr,