Simplify pandas Series of strings that contain missing values

[dfalbel]

Improves string arrays simplification by automatically handling common missing value types from pandas.
The following is now simplified into an R character vector.

pd <- import("pandas", convert = FALSE)
x <- pd$Series(list("a", pd$`NA`, NULL, np$nan))

It's very common for pandas series containg any of pd.NA, None and np.nan and strings otherwise are using those types as missing values indicators. Also, .isna() reports True for all of them.

>   x$isna()
0    False
1     True
2     True
3     True
dtype: bool

[t-kalinowski]

Great work!

I really like this pattern of initializing static PyObject/PyObjectPtr! One downside is that if the first initialization fails, subsequently the function will always throw until the R session restarts, and the error message will not be helpful, since py_fetch_error() won't be able to fetch the Python error a 2nd time. It should be a rare occurrence, though, and easily fixable by restarting the R session.