Read numpy nans/datetimes

[sc1f]

This PR improves the Python API's ability to parse Pandas DataFrames, ensuring compatibility with numpy.nan and numpy.datetime64.

• Columns that begin with numpy.nan will now have their types inferred as float, as numpy.nan is a float type. If the column contains a value (i.e. [nan, nan, 1]) later on, and the value is not of float or integer type, then the column will be promoted to a string, all numpy.nans will be converted to None, and subsequent writes will be converted to string.
• numpy.nan serializes to None when exported from Perspective.
• Timestamp generation has been moved to the date parser, and numpy.datetime64 correctly returns Unix timestamps for use within the C++ engine.

[codecov-io]

Codecov Report

Merging #741 into master will increase coverage by 1.06%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #741      +/-   ##
==========================================
+ Coverage    92.3%   93.36%   +1.06%     
==========================================
  Files          61       61              
  Lines        2703     2970     +267     
==========================================
+ Hits         2495     2773     +278     
+ Misses        208      197      -11

Impacted Files	Coverage Δ
perspective/table/_date_validator.py	86.84% <0%> (-0.66%)	⬇️
perspective/tests/node/test_node.py	31.25% <0%> (ø)	⬆️
perspective/table/view.py	80.95% <0%> (ø)	⬆️
perspective/tests/table/test_table_pandas.py	100% <0%> (ø)	⬆️
perspective/tests/table/test_view.py	98.55% <0%> (+0.09%)	⬆️
perspective/table/_accessor.py	92.13% <0%> (+1.99%)	⬆️
perspective/table/table.py	94.94% <0%> (+3.94%)	⬆️
perspective/tests/table/test_update.py	100% <0%> (+11.23%)	⬆️

---

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 9481128...159af03. Read the comment docs.

[texodus]

Thanks for the PR!