NaN values

Hi team,

I was able to run the charr using hail version. 

```
charr_result = hl.compute_charr(mt, ref_AF=(1 - mt.info.AF[0]))
charr_result.export("charr_result.tsv")
print (charr_result)
```

I am getting values as NaN for most of sample ? How do we interpret this ? Do you have some suggestion for CHAR score to consider as contaminated ?

