[SHUFFLE] manually split of Variable length buffer (String likely)

[FelixYBW]

Is your feature request related to a problem or challenge? Please describe what you are trying to do.
Currently we still use arrow's bufferbuilder to build the binary buffer, which is extreamly not efficient. Each binary builder has 3 memory reference as below. Each reducer each binary column has such a structure needs to be cached in memory.
binarybuilder
offsets_builder_
BufferBuilder
data()
value_data_builder_
BufferBuilder
data()
null_bitmap_builder_
BufferBuilder
data()

Describe the solution you'd like

Solution is to use the data() pointer directly to split the record batch into, just like fixed width columns