submaster: refactor FrequencyTracker to use EMA for simpler and faster frequency calculation

[deanlee]

This update refactors the FrequencyTracker class to enhance its performance and responsiveness by replacing the sum-based frequency calculation with an Exponential Moving Average (EMA). This change significantly improves real-time adaptability and makes the system more efficient in handling dynamic event rates.

Changes:

• EMA-Based Frequency Calculation: Replaces the traditional sum approach with EMA to track frequency. This makes the calculation more efficient and eliminates unnecessary data management.
• Real-Time Responsiveness: The EMA approach allows the class to quickly adjust to real-time changes in event frequency, making the system highly responsive in dynamic environments.
• Less Memory Overhead: By no longer needing to store and sum large historical data, the memory footprint is reduced.
• Improved Performance: The EMA approach avoids the overhead of summing and deque operations, enhancing performance.