Fix the UTF-8 check for partner payload files

Your service ingests files from partners and is moving from Python 2 to Python 3. Python 3 decodes text strictly, so bytes that are not valid UTF-8 raise an error in production. Before the cutover you want to classify each payload the way Python 3 will see it.

An assistant wrote classify_partner_payload(byte_values). It is in the editor. It has a bug. Find it and fix it.

Input and output:

• byte_values: the payload as a list of integers.
• Return "empty", "ascii", "utf8" or "invalid".

Rules the tests check:

• An empty list is "empty".
• If any value is below 0 or above 255, the payload is "invalid".
• Otherwise read the bytes from the start as UTF-8 sequences. A sequence starts with a lead byte:
  • 0 to 127: the byte stands alone.
  • 194 to 223: 1 more byte follows, from 128 to 191.
  • 224: a byte from 160 to 191 follows, then 1 byte from 128 to 191.
  • 225 to 236, 238 or 239: 2 more bytes follow, each from 128 to 191.
  • 237: a byte from 128 to 159 follows, then 1 byte from 128 to 191.
  • 240: a byte from 144 to 191 follows, then 2 bytes from 128 to 191.
  • 241 to 243: 3 more bytes follow, each from 128 to 191.
  • 244: a byte from 128 to 143 follows, then 2 bytes from 128 to 191.
• The payload is "invalid" if a lead byte is any other value (128 to 193, or 245 to 255), if a following byte is out of its range, or if the list ends in the middle of a sequence.
• A valid payload is "ascii" when every byte is 127 or less, else "utf8".