Save event_no as a non-index column in ParquetWriter

[astrojarred]

The main change here is that instead of setting the self._index_column to the index, we just let it stay as a normal column in the dataframe/parquet file.

This way later we don't get 'column not found' errors when merging files in these places:

• _process_shard() when calling id = split[index_column]
• _identify_events() when doing df.select([index_column])

This fixes the bug I was having but would need a maintainer/expert to let me know if this might mess things up downstream, i.e. if somewhere else event_no is expected as an index column. In this case we could just set the index again at the very end of the merging process.

Fixes #871