Refactor FlowMeasure to use dynamic maps and shared locks

[bhagathkrishnacdac]

What type of PR is this?

/kind bug fix

What this PR does / why we need it:
This PR fixes a critical crash in the UPF pod that occurs when the flow_measure module is enabled during UE attachment.
The previous implementation relied on a fixed-size DPDK rte_hash and a separate std::vector for statistics. This architecture was prone to crashes under high-concurrency scenarios (like a burst of UEs attaching) due to:

1. Hash Insertion Failures: If the pre-allocated table filled up or encountered a collision limit, it could lead to unhandled errors in the dataplane.
2. Race Conditions: The locking between the DPDK hash and the stats vector was insufficient, leading to memory corruption or out-of-bounds access during concurrent flow creation.

Changes made to resolve the crash:

• Dynamic Allocation: Switched to std::unordered_map with pointer-based storage (SessionStats*). This allows the module to handle any number of UE sessions dynamically without crashing due to table exhaustion.
• Thread Safety: Implemented std::shared_mutex to properly synchronize the dataplane (shared lookups) and control plane (unique locks for flow creation and table clearing), preventing the race conditions that caused the segmentation faults.
• Safe Fallback: Added a fallback for missing timestamps to ensure that unexpected metadata states do not disrupt the processing logic.

Which issue(s) this PR fixes:

Fixes #32

Test Report Added?:

/kind TESTED

Test Report:

NA

Special notes for your reviewer:
The refactor focuses on moving away from fixed-size DPDK structures to more robust C++ standard containers. This provides the necessary memory safety and locking granularity to handle the bursty nature of UE attachment signaling. The get_or_create logic is the core of the fix, ensuring safe, atomic creation of new flow entries.

[ak2btvm]

Approved