Reset stale process_thread_usage series each scrape

[mempirate]

Summary

• keep the pid and name labels on process_thread_usage
• call reset() before repopulating the gauge vector on each collect
• update docs/examples to reflect the PID-labelled metric output

Why

Keeping pid preserves exact per-thread identity, while reset() prevents stale thread series from accumulating in the exporter after threads exit.

Implementation notes

• GaugeVec::reset() clears the previous scrape's label set before repopulating currently live thread series
• this addresses stale series growth in the exporter, but intentionally keeps PID-labelled per-thread metrics

Testing

• cargo test