fix: add nullability to struct type String()

[Willem-J-an]

Rationale for this change

We frequently run into panic:
panic: arrow/array: column "colName" type mismatch, got=.., want=...
The got and want will be exactly the same schema, the only difference is in the nullability of certain struct fields. The only way to figure out which fields are the problem is to go through the debugger.

The nullability is rendered for other types like map and list, so it seems like it should be there for struct fields as well.

What changes are included in this PR?

• Add nullable in String() representation of StructType fields
• Corrected tests that were affected by this
  • Some were just adding nullable where it's now added to the string and it previously was not
  • Some appear to be actual fixes to the tests, where the schema's did not match, but that was not detected in the schema string comparison because nullability was not included for StructType.

Are these changes tested?

ran make test, didn't have the parquet test files, so didn't run those.

Are there any user-facing changes?

Yes, the String() representation for data types is user facing.

[zeroshade]

Thanks for this! It looks good to me