Collect keys from first 100 rows for arrayDataFrame

[platypii]

Previously it would only look at the first row. This resulted in sub-optimal dataframes with missing columns for data like:

[
  { id: 1, name: 'Alice' },
  { id: 2, age: 25 },
  { id: 3, email: 'charlie@example.com' },
]

This PR changes it to aggregate the keys of up to the first 100 rows of json objects. This should help with jsonl files especially.

[severo]

Makes sense, and it's free since the data is already there.