Refactor ColumnVEPField to be simpler data structures

Now that we have T2T the combinatorial explosion of different builds is starting to bite

Instead of creating multiple entries in the DB, we could have eg ```genome_build = [ "GRCh37", "GRCh38"]``` etc - can do more with JSON than

This is not the final form, but useful to see how much you can remove

```
records = ColumnVEPField.objects.all().order_by("variant_grid_column").values()

cleaned_data = defaultdict(list)
for r in records:
data = {k:v for k,v in r.items() if v is not None}
del data["id"]
if data.get("source_field_has_custom_prefix") is False:
del data["source_field_has_custom_prefix"]
del data["column"]
vgc = data.pop("variant_grid_column_id")
data["variant_grid_column"] = vgc
sf = data.pop("source_field")

cleaned_data[sf].append(data)
```

A few strange things:

Mastermind_counts is processed 3 times (and stored in different fields)
gnomAD_SV_AF is copied to both gnomad_sv_overlap_af and gnomad_sv_overlap_af