pandas2ri - added named-vector-to-pandas conversion

[tikuma-lsuhsc]

This PR adds conversions from R named vector to Pandas Series and R named matrix to Pandas DataFrame.

It resolves the issue I encountered while retrieving the fixed effect coefficients (named vector) from lme4.fixef() and CI matrices from confint() functions.

I have also added test_ri2pandas_named_vector() pytest function to test the changes.

[codecov-commenter]

Codecov Report

Merging #887 (ae76aad) into master (10d6614) will increase coverage by 0.05%.
The diff coverage is 90.90%.

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

@@            Coverage Diff             @@
##           master     #887      +/-   ##
==========================================
+ Coverage   88.97%   89.03%   +0.05%     
==========================================
  Files          39       39              
  Lines        6839     6848       +9     
==========================================
+ Hits         6085     6097      +12     
+ Misses        754      751       -3     

Files	Coverage Δ
rpy2/robjects/pandas2ri.py	90.99% <90.90%> (+0.40%)	⬆️

... and 2 files with indirect coverage changes

📣 Codecov offers a browser extension for seamless coverage viewing on GitHub. Try it in Chrome or Firefox today!

[lgautier]

Hi, thanks for the PR. A review is long due.

I am unsure about whether R named vectors should not be converted to pandas data frames by default, even within the context of the pandas2ri conversion rules. The reason is that while R data.frame objects inherit from R lists, not all R lists can be R data.frames.

Having that said, the current behavior in rpy2 does not seem predictable enough:

import rpy2.robjects as ro
import rpy2.robjects.pandas2ri as pd2r

# First list.

rlist = ro.r('c(a=1, b=2, c=3)')
with (ro.default_converter + pd2r.converter).context() as cv:
    pylist = cv.rpy2py(rlist)

pylist
# array([1., 2., 3.])

# Second list. The conversion is closer to what I would expect, with the exception of
# `b.x` and `b.y` instead of nesting 'x' and 'y' under 'b'. 

rlist = ro.r('c(a=1, b=list(x=list(z=4), y=5), c=3)')

with (ro.default_converter + pd2r.converter).context() as cv:
    pylist = cv.rpy2py(rlist 

pylist
# o{'a': [1.], 'b.x': o{'z': [4.], }, 'b.y': [5.], 'c': [3.], }