We store the sizes as a hash from path to bytes like:

``` ruby
{'ci_artifacts.txt' => 27,
 'other_artifacts_0.1.2/another-subdirectory/banana_sample.gif' =>
71759,
 'other_artifacts_0.1.2/doc_sample.txt' => 1314,
 'rails_sample.jpg' => 35255,
 'tests_encoding/utf8 test dir ✓/regular_file_2' => 7}
```

So that it's easier to access than reading gzip file again.