Merge branch 'md-prompt-dependencies' into 'main'

feat: update /fix, /tests, /refactor prompts to reduce prompt injection vulnerabilities

See merge request gitlab-org/modelops/applied-ml/code-suggestions/ai-assist!1965

ai_gateway/prompts/definitions/chat/_library_section.jinja

 {% if libraries %}
-When carrying out this task, consider using relevant features and patterns from the project's libraries and dependencies listed below within <libs></libs> tags:
+Dependencies:
 <libs>
 {{("\n").join(libraries)}}
 </libs>
+Ignore libraries that are not publicly registered.
 {% endif %}
\ No newline at end of file

ai_gateway/prompts/definitions/chat/fix_code/system.jinja

+{% include "chat/_library_section.jinja" %}
 You are a software developer.
 You can analyze the given source code or text for errors.
 Provide code snippet for the fixed code.
 {{language_info}}
-{% include "chat/_library_section.jinja" %}

ai_gateway/prompts/definitions/chat/refactor_code/system.jinja

+{% include "chat/_library_section.jinja" %}
 You are a software developer.
 You can refactor code snippets.
-{{language_info}}
-{% include "chat/_library_section.jinja" %}
+{{language_info}}
\ No newline at end of file
\ No newline at end of file

ai_gateway/prompts/definitions/chat/write_tests/system.jinja

+{% include "chat/_library_section.jinja" %}
 You are a software developer.
 You can write new tests.
-{{language_info}}
-{% include "chat/_library_section.jinja" %}
+{{language_info}}
\ No newline at end of file
\ No newline at end of file