Add timed pipeline progress logging

[Toti330]

Summary

Adds a time-based progress logger for long-running generation jobs. It reports immediately when generation begins, then reports every 30 seconds by default.

Why this is useful

A long seed-generation run should not be silent. The monitor provides exact generated/committed counts and percentages, interval and average rates, elapsed time, queue occupancy, aggregate worker and writer-pipeline time, likely bottleneck, Linux RSS/peak RSS, and process I/O. This gives a lightweight trace-like view without polling every generation pass.

Implementation

• Adds PROGRESS_LOG=0 to disable the monitor and its timing hooks.
• Adds PROGRESS_LOG_INTERVAL_SECS=N; default is 30.
• Uses a dedicated sleeping monitor thread with a condition variable for prompt final output.
• Wraps worker/writer thread lifetimes rather than timing individual seeds; there is no per-seed timer or instrumentation branch.
• Labels writer timing as writer_pipeline because it includes connection, receive, COPY, and commit work.
• Reads Linux process resource counters from /proc; unsupported systems show n/a.

Independence

This PR is based directly on main and is independently usable. It is compatible with the other feature branches but does not require them.

Verification

• cargo test --all-targets: 3 progress tests passed.
• A 1-second-cadence smoke run reported 36/60 (60.00%) at 1.002s, then emitted a final complete report.
• A PROGRESS_LOG=0 smoke run emitted 0 progress lines.
• Full integration merge with all feature branches passed 8 Rust and 4 Python tests.

---

Generated by GPT-5.6 Terra & Sol. Ideas from the author; implementation by AI.

[SuperB3333]

I'll make sense of this implement some consistency/readability changes and commit it

[SuperB3333]

Implemented 06b6fc3 as a skeleton.
Still need to insert actual values into the log