Refactor data handling

[hechth]

The data handling system could be improved by introducing an internal data representation that handles the data internally as a dataframe and obtains data from readers and exports data using writers. This would allow import and export to use different formats (read from csv, export to msp) and would limit the low-level implementation details such as the presence of keywords etc. to the reader and writer implementations.

# Example
reader = MatchMSReader(".../data.msp", "minutes")
query = Data(reader)

# Access data stored in pandas.Dataframe column "retention_indices" via property
query.retention_indices

# Write data to csv
writer = PandasWriter("output.csv")
query.export(writer)

[hechth]

This could also get substantially easier when matchms/matchms#10 is closed, because simply the spectrum metadata can be used for this, without the spectral data itself, so the internal data format could be purely based on pandas.