- Simplify readFile interface to only support reading full files

- Remove the choice of reading in a blocking way,
  since we want to prevent hanging

- Remove the dry run read, since it was not useful

- Split the implementation between two overloads,
  where one supports returning the whole file contents in one go,
  while the other calls back a predicate everytime some data is read,
  up to a fixed block size (as with the previous implementation)