export_csv drops fields that are missing from the first record

[chiruu12]

export_csv builds the column list from the first record only:

if fields is None:
    fields = list(data[0].keys())

Rows are then built by looking up fields on each record, so a key that appears on any later record is never written and nothing is raised.

from brightdata.datasets.utils import export_csv

data = [
    {"url": "a.com", "price": 10},
    {"url": "b.com", "price": 20, "discount": "50%"},
]
export_csv(data, "out.csv")

out.csv:

url,price
a.com,10
b.com,20

discount is gone with no warning. Scraper results are commonly heterogeneous, since optional fields are absent when a page does not have them, so this loses real data on ordinary output. export_json and export_jsonl keep every key, so the same result set exports differently depending on format.

Taking the union of keys across all records, in first-seen order, would fix it. Happy to send a PR.