fix data structure for nebula datasource

[Nicole00]

No description provided.

[caton-hpg]

1.May be simpler code:

      var df = spark.read
          .nebula(config, nebulaReadEdgeConfig)
          .loadEdgesToDF()
        if (weights.nonEmpty) {
          df = df.select("_srcId", "_dstId", weights(i))
        }

      dataset = dataset == null ? df:dataset.union(df)

2. There may be wrong when hasWeight is false and weights is nonEmpty.
   Because there is not the weights(i) column in df.
   But There are right when it executes.
   Because It will fetch all props of edge when hasWeight is false.

3. There will be java.lang.IndexOutOfBoundsException: 1 if the num of labels does not equal the num of weights.
   It's a just configuration error.

[Nicole00]

3. if the num of labels does not equal the num of weights

1. yeah, code will be unified.
2. If hasWeight is false and weights is nonEmpty, the graph will has default 1.0 weight, means df has three columns and the third column has the same value 1.0. When hasWeight is false, the code will fetch no props of edge, because the config withNoColumn is true.
3. good catch, will add config check.