Made maximum number of output tokens configurable, and added debug output.

[lukehollenback]

Just in case https://github.com/freeedcom/ai-codereviewer/pull/57 does not fully get to the bottom of the JSON parsing issue in Issue #56, this pull request adds some extra debug output.

It also makes the maximum number of output tokens configurable. Another suspicion is that the hardcoded 700 limit was just too small for some large pull requests that needed a lot of commentary.