Fix the streaming row counter for S3 CSV uploads

Your Lambda counts the rows in a CSV file that lands in S3. The first version read the whole file into memory and split it into lines, so real uploads of a few gigabytes ran out of memory. The new version reads the object as a stream of chunks and keeps only a few counters.

An assistant wrote count_csv_rows_streaming(chunks, has_header). It is in the editor. It has a bug. Find it and fix it.

Input and output:

• chunks: the file's text as a list of strings, in order. The file is all the chunks joined together. A chunk can be empty, and a chunk can end anywhere, even in the middle of a row.
• has_header: true when the first row of the file is a header row.
• Return the number of data rows, as an integer.

Rules the tests check:

• Only \n ends a row. Every other character is part of a row, \r included.
• A file with no characters has 0 rows.
• Otherwise, the number of rows is the number of \n characters in the file, plus 1 when the file's last character is not \n.
• So a final \n does not start an extra empty row, but every other \n ends a row, even when the row is empty.
• When has_header is true and there is at least 1 row, the first row is the header and is not counted.