[Feature request] Support nodes transform and with multiple textual fields

Some nodes may contain more than one fields of texts (e.g. title, key words, description). The current best practices is to concatenate them before tokenization. However, there are cases where concatenation is infeasible: leading to long texts beyond LM's receptive fields; and mixed types of languages:

```
{
"node_type": "drug",
"format": {
"name": "parquet"
},
"files": [... ],
"node_id_col": "ent idx",
"features": [
{
"feature_col": "name",
"feature_name": "name",
"transform": {
"name": "tokenize_hf",
"bert_model": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext",
"max_seq_length": 256
}
},
{
"feature_col": "SMILES_string",
"feature_name": "SMILES_string",
"transform": {
"name": "tokenize_hf",
"bert_model": "DeepChem/ChemBERTa-77M-MLM",
"max_seq_length": 256
}
}
]
}
```

No files, tests, or entry points are named in the issue. Start by locating the existing node feature transformation and tokenization configuration, then determine how multiple textual feature fields can retain separate transforms; done means fields such as name and SMILES_string can be processed independently without concatenation.

Written by the indexing model from the issue text.