0.8.2

• Fix primary model name selection logic to prefer names with longer tags.
• Propagate input/output token usage information to llm.
  To see token usage, specify -u option, e.g.:

  $ llm -u -m llama3.2 "How much is 2+2?"
  The answer to 2 + 2 is 4.
  Token usage: 33 input, 13 output