AccumulatorContext is a
private[spark] internal object used to track accumulators by Spark itself using an internal
originals lookup table. Spark uses the
AccumulatorContext object to register and unregister accumulators.
originals lookup table maps accumulator identifier to the accumulator itself.
Every accumulator has its own unique accumulator id that is assigned using the internal
AccumulatorContext.SQL_ACCUM_IDENTIFIER is an internal identifier for Spark SQL’s internal accumulators. The value is
sql and Spark uses it to distinguish Spark SQL metrics from others.