fix(ignore): avoid nanosleep busy loop in worker fallback (#2761)

[krapcys1-maker]

Fixes #2761

Summary

Fixes a busy-loop fallback in ignore walker worker scheduling that could spin in nanosleep while another worker is blocked on a FIFO.

Problem

When one worker blocks on a FIFO, other workers can enter a repeated nanosleep cycle in Worker::get_work instead of yielding cleanly.

What changed

• Replace millisecond sleep fallback in Worker::get_work with scheduler yield (yield_now).

Solution

Use std::thread::yield_now() in the non-blocking stack fallback path so workers stop burning CPU while waiting for progress.

Why

This preserves behavior while reducing unnecessary syscall/CPU churn in stalled traversal scenarios.

Tests

• cargo test -p ignore walk -- --nocapture
• cargo test -p ripgrep --tests