Fix #1: reduce CPU usage during SMB file copy monitoring

[elstefan-assistant]

Summary

This PR addresses #1 (high CPU load during SMB copy) by removing per-packet heavy allocations/parsing and throttling expensive fallback paths.

Changes

• PackageWorker
  • Replaced full-buffer copies with exact-length pooled buffer copies via ArrayPool.
  • Removed repeated receive buffer re-allocation and now reuses a fixed packet buffer.
  • Added fast-path packet parsing for IPv4/TCP destination port 445 (SMB) without allocating full IpHeader/TcpHeader objects for every packet.
  • Returns pooled buffers in finally to avoid memory growth/leaks.
• WolHelper
  • Throttled unknown-IP logging (max once per IP per 5 minutes) to reduce log I/O pressure.
  • Throttled ARP refresh attempts (max once per minute globally) instead of refreshing on every unknown packet.

Expected Impact

• Lower CPU utilization under heavy SMB traffic due to far fewer allocations and less object construction per packet.
• Lower disk I/O from reduced repeated unknown-IP logs.
• Preserved WoL behavior for SMB destination traffic (dst port 445) with existing wake cooldown logic intact.

Validation

• Static behavior validation completed from code-path review (port-filtering and wake path preserved).
• Runtime benchmark/functional verification requires Windows host + .NET SDK/runtime due to raw socket + Windows service constraints noted in README.

Closes #1