Remove super().assert_extra_args() call in sdxl_train

[MakkiShizu]

Description

Fixes AssertionError when training SDXL LoRA models with bucket_reso_steps=32.

Problem:

• Parent class train_network.py #L98 enforced

    def assert_extra_args(self, args, train_dataset_group):
        train_dataset_group.verify_bucket_reso_steps(64)

• SDXL subclass attempted to set 32 after parent validation, causing conflict

Changes:

1. In sdxl_train_textual_inversion.py,sdxl_train_network.py Removed super().assert_extra_args() in SdxlNetworkTrainer.assert_extra_args

Verification:

• Confirmed no more bucket_reso_steps=32 assertion failures in SDXL mode

[kohya-ss]

Thank you for this, and I apologize that I did not notice this PR. I merged the same update in #1966, so I will close this one.