Fix O(n^2) complexity in rule check and group creation

[Swapnil-jain]

Summary

• Fixes There is an O(n^2) logic in rule check. (n = number of questions) #2376
• Optimizes O(n²) complexity to O(n + E) in rule check and group creation methods

Root Cause

Three locations had inefficient nested loops:

1. add_stop_rule in rule_manager.py - loop continued even after finding a match
2. suggest_dependency_aware_groups in survey.py - nested iteration through group members for each candidate
3. create_allowable_groups in survey.py - same nested iteration pattern

The Fix

1. add_stop_rule: Changed to any() which short-circuits on first match
2. suggest_dependency_aware_groups & create_allowable_groups:
   • Pre-compute an inverted DAG (inverted_dag[j] = questions that depend on j)
   • Use set intersection instead of iterating through all group members
   • Maintain current_group_set incrementally instead of recreating each iteration

Complexity Improvement

Method	Before	After
add_stop_rule	O(n) always	O(1) to O(n) with early exit
suggest_dependency_aware_groups	O(n² × |group|)	O(n + E)
create_allowable_groups	O(n² × |group|)	O(n + E)

Where n = number of questions, E = number of dependency edges

Test Plan

• Existing tests in tests/surveys/test_group_navigation.py cover the modified methods
• Logic verified to be equivalent to original implementation

[Swapnil-jain]

Would appreciate a review. @johnjosephhorton @zer0dss @rbyh