Use _call within _get_file to ensure retries

[oliverwm1]

Use _call within _get_file to ensure that the retry logic in that function is exercised. This PR makes a significant change in implementation, since now multiple http requests are made within single _get_file call (if the file is large) whereas previously a single request was made. Not sure about performance implications (speed, memory usage, reliability).

Resolves #383

[martindurant]

@isidentical , this is probably also of interest to you

[oliverwm1]

@martindurant @isidentical thoughts on this? It passes the (skipped) test_get_put tests for me locally, and I've manually called fs.get on a small and large file with success. I'm not sure about performance implications and also perhaps there are complications for compressed files? (https://cloud.google.com/storage/docs/transcoding#decompressive_transcoding)

[martindurant]

I'm pretty sure we need a way to split up the reads and stream to disc, else we risk memory issues.

[oliverwm1]

I'm pretty sure we need a way to split up the reads and stream to disc, else we risk memory issues.

Okay. Other idea is to define a retry decorator that we can apply to _call and _get_file. Not totally straightforward to get an interface that will work for both functions, but I think it's doable. Trying to avoid more copy and paste.

[martindurant]

Perhaps you can pass the output file into _call ?

[oliverwm1]

Closing in favor of #387