[BUG] model_merger sets lora_alpha=0 when merging SFT LoRA checkpoints

[Yatogaii]

System Info

No relevant with system info.

Information

• The official example scripts
• My own modified scripts

Tasks

• An officially supported task in the examples folder (such as GLUE/SQuAD, ...)
• My own task or dataset (give details below)

Reproduction

1. Train an SFT model with LoRA enabled, for example:

model.lora_rank=8 
model.lora_alpha=16

2.Merge the checkpoint using:

python3 -m verl.model_merger merge

3. Inspect the generated LoRA config in the merged output directory.
   The merged adapter_config.json contains:

    "lora_alpha": 0,

Expected behavior

lora_alpha in the merged checkpoint should match the value specified during training (e.g., 16 in this example).

As mentioned in #3050, this issue still exists in the current version.
This behavior effectively makes SFT + LoRA workflows unusable after merging, since the scaling factor is lost.