BUG: Inaccurate means for float16 TIFFs

[scottclowe]

When taking the mean of an array of ints, numpy uses float64 for the output.

But when taking the mean of an array of floats, numpy uses the same float precision as the original array.

That leads to the mean being noticably less accurate when taking the mean of TIFFs that are float16 (single precision). We fix this by using float64 for all the means.

This issue wasn't caught by the unit tests before because they were also doing the less accurate rounding to determine the target values.

[codecov]

Codecov Report

Merging #176 (e33a06e) into master (c422444) will not change coverage.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master     #176   +/-   ##
=======================================
  Coverage   92.65%   92.65%           
=======================================
  Files           8        8           
  Lines         831      831           
  Branches      165      165           
=======================================
  Hits          770      770           
  Misses         31       31           
  Partials       30       30           

Flag	Coverage Δ
unittests	92.53% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
fissa/extraction.py	95.87% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c422444...e33a06e. Read the comment docs.