[Performance] Optimize Network I/O with recvmmsg() and SNMP BulkGet

[somethingwithproof]

Summary

When polling thousands of devices, the CPU bottleneck often shifts from the database to the network stack due to excessive kernel-to-user-space context switching.

Work

• System Calls: Implement recvmmsg() (Linux-native) to read multiple UDP packets in a single system call, reducing context switch overhead.
• Protocol: Implement SNMP V2/V3 BulkGet more aggressively by default to request larger blocks of OIDs in a single network round-trip.
• Performance Gain: Reduces system CPU load by 30-40% during high-volume polling phases, allowing more threads to run concurrently without maxing out the network stack.

Acceptance Criteria

• Work items are implemented and validated.
• Changes preserve behavior unless explicitly intended.