test: add comprehensive benchmark suite

[taigrr]

Fixes #6

Changes

• Added log/bench_test.go with 13 benchmarks covering:
  • Client creation/destruction overhead
  • All log levels (Trace, Debug, Info, Warn, Error)
  • Formatted logging (Tracef, Infof)
  • Multiple client fan-out (5 consumers)
  • Parallel concurrent writes
  • Namespace filtering performance
  • Component-level: fileInfo (runtime.Caller) and entry creation baselines

Key Findings

• ~1.0-1.2μs per log entry (single client, single goroutine)
• ~2.6μs with 5 client fan-out
• runtime.Caller accounts for ~65% of per-entry cost
• Namespace filtering saves ~20% when messages are filtered out
• 6 allocs/op for standard log calls, 7 for formatted variants

Testing

go test ./log/ -bench=. -benchmem