[Feature request] `readbytes(path, start, stop)` to efficiently read a byte range from a file

[asinghvi17]

Accessing chunked data often requires reading specific, pre-known byte ranges from a file.

S3 already supports this (AWSS3.jl has a keyword to read for this reason) and some HTTP implementations also support retrieving a certain range of bytes from the response.

The ideal function would have a generic implementation that looks like:

function readbytes(path, start, stop)
	open(f) do io
  	  seek(io, start)
 	  read(io, stop-start)
	end
end

and packages like AWSS3.jl could override if they have more efficient ways to perform this kind of access.