Error saving series with all missing values

## Bug Report

### Steps to Reproduce:
In an IDE using the tracdap runtime try to save a dataFrame as an output of a model:

Create the dataFrame with a column made up completely of missing values:

```
data = {
'INDEX': ['A', 'B', 'C'],
'INDEX_VALUE_DATE': [date, date, date],
'INDEX_VALUE': [pd.NA, pd.NA, pd.NA]
}

df = pd.DataFrame(data)
```

With the schema:

```
field_name,field_type,label,categorical,business_key,format_code
INDEX,STRING,Index name,FALSE,FALSE
INDEX_VALUE_DATE,DATE,Index value date,FALSE,FALSE,"DAY"
INDEX_VALUE,FLOAT,Index value,FALSE,FALSE,"|.|3|||1"
```

### Expected Result:
Data is saved as an output without error.

### Actual Result:
Error occurs

```
raise _ex.EModelExec(msg) from e
tracdap.rt.exceptions.EModelExec: There was an unhandled error in the model: Column 2 named INDEX_VALUE expected length 3 but got length 0
```

### Environment:

Python 3.11
tracdap runtime 0.5.9

### Additional Context:
Saving the data works if one value in the column is not NA.

No repository file or test is named in the report. Reproduce the failure in the Python 3.11 runtime with the shown pandas DataFrame and schema, then trace the output-saving path for the all-NA INDEX_VALUE column. Done means the output saves successfully with three rows, matching the behavior when one value is non-missing.

Written by the indexing model from the issue text.