[3.12] gh-119118: Fix performance regression in tokenize module (GH-119615) (#119683)

- Cache line object to avoid creating a Unicode object
  for all of the tokens in the same line.
- Speed up byte offset to column offset conversion by using the
  smallest buffer possible to measure the difference.

(cherry picked from commit d87b0151062e36e67f9e42e1595fba5bf23a485c)

Co-authored-by: Lysandros Nikolaou <lisandrosnik@gmail.com>
Co-authored-by: Pablo Galindo <pablogsal@gmail.com>