configurable max data repairs

[pelikhan]

• The code change primarily involves modifications to functions applyRepairs and processChatMessage in chat.ts.
• The trace message has been updated in the applyRepairs function to say "data repairs" instead of just "repair" 🔧.
• The formatting of the repair message constructed in the applyRepairs function has been changed, with more emphasis on the importance of repairing formatting issues 🧹.
• The processChatMessage function now directly utilizes the maxDataRepairs option from the function's arguments, instead of reassigning it to a local variable. This may help reduce memory usage and increase efficiency 🚀.
• A throw error in processChatMessage has been restructured for better readability 📖.
• The maxDataRepairs option has been added to the ModelOptions interface in prompt_template.d.ts. This could be to limit the number of data repairs a model can attempt, adding a control mechanism ⚙️.
• Added a comment to the DefSchemaOptions interface describing the format field 📝.

generated by pr-describe

[github-actions[bot]]

The changes in the pull request appear to be primarily around message formatting and the handling of data repairs within the chat functionality.

1. The repair message formatting has been changed to use a dedent template literal with some modified text.
2. The maximum number of data repairs allowed (maxDataRepairs) is now also an option that can be set when calling processChatMessage, falling back to a default if not set.
3. A comment has been added in the definition file to clarify that format refers to the output format in the prompt.

Concerns: No major functionality concerns. However, make sure that the consumers of the processChatMessage function are aware of the new maxDataRepairs option and its implications.

LGTM :rocket:

generated by pr-review