Follow-up from "Improve performance of writing live trace chunks"

The following discussion from !37075 (merged) should be addressed:

• @ayufan started a discussion:

  I think we could optimise a number of requests to do something like this:

        @size ||= @data&.bytesize || current_store.size(self) || data&.bytesize

  Effectively:

  1. try cached data
  2. use efficient size
  3. read data and get size