Explicitly test for nan being handled correctly

[shntnu]

No description provided.

[diskontinuum]

This was partly addressed by the latest merged PR.
If the files contain NaNs, then the easiest way to ingest them would be:

1. use the Parquet backend by adding the flag --parquet
2. modify the config.ini file that is used in the command: In the [schema] section set type_conversion = all2string.

The ingested file contents are now all stored as type string. This might be a problem downstream, so this is generally not the preferred setting. However, conversion to str might be the last resort in a situation where the files have different types and NANs for columns that will be concatenated and are tied to a schema type.