stream the file rather than store it in memory

from discussion in https://www.reddit.com/r/ruby/comments/4bcgjw/download_url_to_tempfile_gem/

tips from Reddit user `moomaka`

> This implementation downloads the entire file into memory then writes it to disk. This is not only a bad idea for performance but is also a DoS vector.

Implementation recomendation

> Probably the most efficient is IO.copy_stream or similar methods from IO: http://ruby-doc.org/core->2.3.0/IO.html. Another option is Net::HTTP streaming responses: http://ruby-doc.org/stdlib->2.3.0/libdoc/net/http/rdoc/Net/HTTP.html#class-Net::HTTP-label-Streaming+Response+Bodies
> Regardless of which approach you use, you probably want to set an upper bound on the size of the >file also or someone could just point you at a 15TB file :)