Make hardware tuning and estimates easier to understand

[chrisch88dev]

Summary

• add evaluate and guess commands for hardware checks and pattern estimates without generating keys
• add low, medium, and full power modes with measured worker and batch selection inside each CPU cap
• replace abbreviated throughput and candidate terminology with plain descriptions of possible addresses tested per second
• explain average, 50 percent, and 95 percent estimates in CLI output
• group exact worker, batch, and timing controls under advanced help
• improve sub-millisecond ETA formatting and document the simpler workflow

Validation

• cargo fmt -- --check
• cargo clippy --all-targets --all-features -- -D warnings
• cargo test --all-targets
• cargo audit
• cross-target checks for Windows GNU, Intel macOS, and Apple Silicon macOS
• release-mode evaluate, guess, benchmark, search, key save, and key verification smoke tests

A full-power 500 ms release smoke benchmark selected 10 workers with batch 1024 at 16.73 million possible addresses per second on the current development machine. This is recorded here as validation context, not a general performance claim.