[Intermediate Representation] XGBoost codegen refactor TODO list

[tonyyang-svail]

As discussed in the design doc #785, we will generate the Python submitter program from the intermediate representation.

1. Transfer the training logic to TrainIR.
   1. Generate the training program from mocked TrainIR. [Intermediate Representation] Generate the XgBoost training program #827
      1. Create a Python database connector from the data source string. [Python] Generate Python database connector from data source string #806
   2. Generate TrainIR from SQL
      1. Unify feature column and column spec. Unify featurecolumn and column spec #799
      2. Make expression resolver realize type. resolveExpression should distinguish string, integer and float #800
   3. Enable SQLFlow to execute the XGB train code generated by IR. [Intermediate Representation] Enable SQLFlow to execute the XGB train code generated by IR #980
      1. Use pkg/sql/codegen/attribute in XGBoost TrainIR. Use pkg/sql/codegen/attribute in XGBoost TrainIR #1002
      2. Add XGBoost training documentation. Add XGBoost training documentation #1003
2. Transfer Analysis to AnalyzeIR. [Intermediate Representation] Analysis codegen refactor TODO list #978
3. Transfer the predict logic to PredictIR. [Intermediate Representation] XGBoost predict using IR #1051

[typhoonzero]

After we finish above work, we should add a validator to validate attributes and generate documents for each codegen models: #774