Add Form Context to the Prompts

[parsaalian]

To know what the form is about in the first place, we can use aggregate all the labels in the form, plus the global feedback, and use it in the prompt.

[parsaalian]

Added labels in the prompt as the form context.

Changes in system prompt for constraint generation:

The information that user has provided is as follows:
- All the labels in the form, so that you would have a context of what the form is about.
- The input and the textual information associated with it, such as labels, hint texts, or inline feedback.
- The input fields that are relevant to the input field in question.
- Previously generated constraints for the input field in question if available.
- Tested values and the inline feedback for the input field in question if available.
- Test values and the global feedback for the whole form if available.

Changes in user prompt for constraint generation:

The labels for the form are:
{form_context}
These labels are used to provide context for the functionality of the form.

We are generating constraints for the following input field:
{create_field_info_text(input_group)}

The relevant information available in the form are (in order of relevance):
{create_relevant_info_text(input_group, relevant_count)}

{"The previously generated constraints are:" if generated_constraint_string is not None else ""}
{generated_constraint_string if generated_constraint_string is not None else ""}

{"We have tried to generate constraints for this field before." if last_try is not None else ""}
{last_try["value"] if last_try is not None else ""}
{"And got the following inline and global feedback:" if last_try is not None else ""}
{last_try["feedback"] if last_try is not None else ""}

[nashid]

The prompt appears to be quite verbose, taking up approximately 550 tokens. I assume this is intentional.