`/changelog summarize` should not consume tokens

Since the summary is not user-specific, the responses can be cached and reused. This will also improve latency

Start at the `/changelog summarize` command implementation and trace how it requests the summary and accounts for token use. Check how responses could be reused for non-user-specific input; done means repeated summaries use the cached response, consume no tokens, and retain the expected output while improving latency.

Written by the indexing model from the issue text.