`read_values` (and `read_pandas`) fail on large data or large dataframe

We are getting, 

```
Job failed with error RuntimeError: UDF function/class data is too large!(view logs)
```

for the code like this:

```python
data_sample = parsed_df
input_df = dc.read_pandas(data_sample).select("test")
```

if data is large enough (>10M limit)

We need to rewrite `read_values` to avoid using `gen` and do direct operations.