Stream downloads to save memory

[McToel]

This addresses #1016.

The changes were way simpler than I thought, but it only saves about 33% of memory. That's still quite good, I think.

Due to the streaming, I had to remove a log message, as it is no longer possible to count the downloaded data before consuming it.

Any suggestions for code improvements are welcome, and I have not run any benchmarks yet, on how much memory it is saving. The 33% come from looking at the memory usage of Python in my system monitor.

[gboeing]

@McToel looks like tests are failing.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.06%. Comparing base (f38ddd1) to head (d48bd72).

Additional details and impacted files

@@            Coverage Diff             @@
##              mem    #1021      +/-   ##
==========================================
- Coverage   98.07%   98.06%   -0.01%     
==========================================
  Files          24       24              
  Lines        2439     2430       -9     
==========================================
- Hits         2392     2383       -9     
  Misses         47       47              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.