Fix column names in df upload v2

[simon-contreras-deel]

fix: #922

In addition to the issue info, I am solving another problem: until now, we were filtering column names, getting the geometry and getting column names in different places (table creation, copy query, _rows, ...) As a result, we could get a different number of columns in different places. Now, we are going to use DataframeColumnsInfo to get all the info once at the begining.

[simon-contreras-deel]

There are several changes, so it will need acceptance in addition to the CR

[alrocar]

Nice tests for a cumbersome problem.

I left a couple of comments and a question.

Feel free to merge this as it is.

[alrocar]

Wait, I leave acceptance for someone else.

Maybe @esloho can help with acceptance.

[simon-contreras-deel]

Tips for acceptance:

• check the issue Dataset(df).upload() doesn't work if the only columns are the_geom and cartodb_id #922 and the description
• check the notebooks and examples related with dataframe Datasets.
• other related issues: Error uploading dataframes #947 Dataset().upload() does not work without a geopandas dataframe #914

[Jesus89]

OK LGTM. It seems that the upload is working 🎉 I have tested the publication too.

I have some recommendations:

• I have tried incorrect values of if_exists, like if_exists="replacex", and there is no error with the available values of the parameter. It would be nice if we have some error or warning to clarify the user.
• I have published a map with a SQL + Maps API key. I would expect also a warning message telling the user that this API key has also SQL permissions or access to other datasets. I don't know if this is possible or not, but it would be nice to avoid security issues.

[simon-contreras-deel]

Ok cool. I am going to merge it. About the suggestions, let us discuss them in next iterations. Thanks