Multisheet load/save consistency in ExcelDataSet

[sbrugman]

The PR [#867] introduces a new class MultisheetExcelDataSet that is able to write multiple sheets to an excel file. The load functionality is not implemented. As an alternative it is suggested to load each sheet separately while specifying each sheet name.

According to the pandas documentation read_excel, which is used in ExcelDataSet, returns a dictionary of all sheets if parameter sheet_name is set to None. It's also possible to specify a list of sheet names, returning a dictionary of that subset of sheets.

This means that strictly speaking, the return type of this function is Union[pd.DataFrame, Dict[str, pd.DataFrame]] (assuming str sheet names, as per pandas docs).

This introduces inconsistency between the save and load functions of the ExcelDataSet. A logical solution is to support writing multiple sheets to excel to mitigate this. For this, PR [#867] should be modified to incorporate the MultisheetExcelDataSet into the ExcelDataSet. Introducing this new class is then obviated.

[antonymilne]

Done in #963, thank you very much!