refactor: deduplicate VA-API decode loops between vaapi_h264.rs and vaapi_av1.rs

[staging-devin-ai-integration]

Summary

vaapi_h264.rs and vaapi_av1.rs share ~200 lines of nearly identical decode loop / drain_decoder_events logic, including the EAGAIN_YIELD_THRESHOLD backoff pattern. Similar patterns also exist in dav1d.rs and av1.rs.

Suggestion

Abstract the common decode loop into a generic vaapi_decode_loop<C>() function (or similar) to avoid maintaining multiple copies of the backoff logic in sync. This would reduce the risk of divergent bug fixes.

Files involved

• crates/nodes/src/video/vaapi_h264.rs — drain_decoder_events, EAGAIN_YIELD_THRESHOLD
• crates/nodes/src/video/vaapi_av1.rs — drain_decoder_events, EAGAIN_YIELD_THRESHOLD
• crates/nodes/src/video/dav1d.rs — similar EAGAIN_YIELD_THRESHOLD pattern
• crates/nodes/src/video/av1.rs — similar EAGAIN_YIELD_THRESHOLD pattern

Context

Identified during code review of PR #279 (custom VA-API H.264 encoder shim). Scoped out of that PR to keep changes focused.

[staging-devin-ai-integration]

Closing — the VA-API decode loop deduplication (the main concern) was addressed in PR #279, which extracted a shared vaapi_decode_loop_body<D>() function used by both vaapi_h264.rs and vaapi_av1.rs.

The remaining scope (dav1d.rs and av1.rs still having independent EAGAIN_YIELD_THRESHOLD patterns) has been filed as a narrower follow-up in #302.