CI: Virtualenvs in AppVeyor

[effigies]

This place is not a place of honor.

This is an experiment along the lines of #676, #677, #680 #696.

[codecov-io]

Codecov Report

Merging #745 into master will decrease coverage by 1.66%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #745      +/-   ##
==========================================
- Coverage   89.57%   87.91%   -1.67%     
==========================================
  Files          93      186      +93     
  Lines       11470    22940   +11470     
  Branches     1991     3982    +1991     
==========================================
+ Hits        10274    20167    +9893     
- Misses        856     2097    +1241     
- Partials      340      676     +336

Impacted Files	Coverage Δ
nibabel/environment.py	75% <0%> (-20%)	⬇️
nibabel/casting.py	85.59% <0%> (-1.7%)	⬇️
nibabel/nicom/dicomwrappers.py	90.73% <0%> (-0.82%)	⬇️
nibabel/dft.py	80.48% <0%> (-0.61%)	⬇️
nibabel/nifti1.py	91.17% <0%> (-0.31%)	⬇️
venv/Lib/site-packages/nibabel/tripwire.py	100% <0%> (ø)
venv/Lib/site-packages/nibabel/cmdline/nifti_dx.py	44.44% <0%> (ø)
venv/Lib/site-packages/nibabel/brikhead.py	97.57% <0%> (ø)
...nv/Lib/site-packages/nibabel/streamlines/header.py	100% <0%> (ø)
venv/Lib/site-packages/nibabel/dft.py	15.85% <0%> (ø)
... and 88 more

---

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 abc2f43...c24d1b1. Read the comment docs.

[effigies]

Well, tests pass, so it at least won't cause harm. Perhaps the issue was somehow two copies of numpy somehow getting loaded and compared against each other. Merging, in hopes that the stochastic failures will go away.

[effigies]

Abandon all hope. https://ci.appveyor.com/project/nipy/nibabel/builds/23292543/job/0ypumy06fg2dsjan