allow for caching subsets of data

[gidden]

For reporting subsets of data (e.g., only one region), we will not want to pull all data into memory. However, the caching system currently only supports that. Key lines are:

    def element(self, ix_type, name, filters=None, cache=None):
        """internal function to retrieve a dataframe of item elements"""
        item = self.item(ix_type, name)

        # if dataframe in python cache, retrieve from there                                                                                                                       
        if name in self._pycache:
            return filtered(self._pycache[(ix_type, name)], filters)

        # if no cache, retrieve from Java with filters                                                                                                                            
        if filters is not None and not self._cache:
            return _get_ele_list(item, filters, **self._java_kwargs[ix_type])

        # otherwise, retrieve from Java and keep in python cache                                                                                                                  
        df = _get_ele_list(item, None, **self._java_kwargs[ix_type])

        # save if using memcache                                                                                                                                                  
        if self._cache:
            self._pycache[(ix_type, name)] = df

        return filtered(df, filters)

Here we would need to update how it works if cache is True and filters exist. Probably very low priority.

[khaeru]

CachingBackend, added by #213, can cache a subset of all data in an item. See also #222 for follow-up discussion.