More robust import error handling

[MalteJanz]

Currently if a sync API request fails for import:

• It checks the error body for "data pointers"
• Removes these invalid rows from the chunk and reports them to the user (with the line number in the CSV file)
• Retries with another final request

There should also be a more general error handling if the above doesn't work:

• write the entities one by one, also report errors to the user

See src/data/import.rs

[MalteJanz]

Also the current implementation may actually be wrong. There could be multiple errors for a single line and in that case it might actually remove some valid lines too 🙈

See:

sw-sync-cli/src/data/import.rs

Line 115 in c07d7f8

fn remove_invalid_entries_from_chunk(

It stores to_be_removed in a Vec, so it could contain the same index twice. Maybe we should use a HashSet instead or something else to ensure the indices are unique