test: add benchmark suite for scan and export commands

[mendsec]

Context

catnet currently has no benchmark tests. Because catnet is a thin wrapper, the relevant benchmark is end-to-end CLI invocation latency (binary startup + scan of localhost + JSON export). Establish a baseline so regressions introduced by new flags or middleware are detectable in CI.

Acceptance criteria

• BenchmarkScanLocalhost measures full StartScan of 127.0.0.1 with ports 80,443 via the engine directly (not subprocess)
• BenchmarkExportJSON measures ExportJSON of a 1000-device ScanReport (mirrors catnet-core benchmark for comparison)
• go test -bench=. -benchmem ./... runs without error in CI
• Baseline numbers documented in docs/performance.md