[Code scan] Fix none embedding constructor and node output property

[njzjz]

This issue is part of a Codex global repository scan.

Problem:
The Embedding(method="none") implementation calls torch.nn.Module.__init__ with extra arguments. It also exposes out_note_dim instead of out_node_dim, while NNENV reads self.embedding.out_node_dim when building SKTB prediction layers.

Code references:
https://github.com/deepmodeling/deeptb/blob/86c60c73996f0dd961c3138f2e88424382cb734e/dptb/nn/embedding/identity.py#L8-L24
https://github.com/deepmodeling/deeptb/blob/86c60c73996f0dd961c3138f2e88424382cb734e/dptb/nn/deeptb.py#L131-L143

Impact:
The none embedding path fails during model construction, and would still fail later with AttributeError even after the constructor is fixed.

Suggested fix:
Call super().__init__() without extra arguments and rename/add the property as out_node_dim.