• Remove ref. to class doctsring: parameters are new documented in
  method docstring.
• Don't fill the "bad" values in the array to avoid a warning; correctly
  set the warnings filter instead (hopefully should be faster, for large
  inputs).
• Don't compute np.isnan(X) twice -- once is enough. Don't bother
  "optimizing" by checking whether np.any(np.isnan(X)) -- that'll need
  iterating over the whole array as well.
• Remove reference to very old (and now fixed) putmask endianness bug. (fast putmask implementation does not work on big-endian systems (Trac #575) numpy/numpy#1173)
• Indexing is just as fast as ndarray.take nowadays, and clearer. (np.take and fancy indexing return arrays with different layouts numpy/numpy#9450 (comment) -- np.take use was added for performance in 31deac2)

PR Summary

PR Checklist

• Has Pytest style unit tests
• Code is Flake 8 compliant
• New features are documented, with examples if plot related
• Documentation is sphinx and numpydoc compliant
• Added an entry to doc/users/next_whats_new/ if major new feature (follow instructions in README.rst there)
• Documented in doc/api/api_changes.rst if API changed in a backward-incompatible way