[Offline Pipeline] Batch processing

[sarahwooders]

Offline processing can be made more efficient by batch-processing data, so we should offer an option to define a batch update function for the operators.

• user-defined on_records(self, records: List[Record]) → List[Record]
• set batch size for operators

class Operator: 

  def __init__(self, batch_size): 
    self.batch_size = batch_size

  # user defined transformation   
  def on_record(self, record: Record): 
      pass 

  # can be overwritten by user 
  def on_records(self, records: List[Record]): 
    for record in records: 
      self.on_record(record)