fix(remote): bound sync readiness by wall clock

[pennywise1026-eng]

What changed

Bound remote.sh sync start readiness polling by both its existing attempt ceiling and a wall-clock budget. The defaults preserve the documented approximately 16-second wait; AGMSG_SYNC_READY_TRIES and AGMSG_SYNC_READY_SECONDS permit explicit test control.

Why

The previous 1600-iteration loop also ran process status, tail, awk, and sleep on every turn. On slow Windows/Git Bash filesystems it could take roughly two minutes rather than the documented wait. A late or missing ready marker should cost only this caller its intended time budget.

Validation

• Added a handshake regression that slows both post-start status routes, then proves timeout cleanup is reached before an external lock holder can be retaken.
• Removing only the wall-clock condition makes that regression fail within its bounded wait; restoring it passes.
• git diff --check and bash -n scripts/remote.sh pass.
• Focused Lock and startup retry budgets are counted in iterations, so their stated timeouts hold only on POSIX-speed filesystems #779 and existing the Windows lock was never left behind: a live sync start holds it while polling readiness #817 retake-cleanup tests pass.

Refs #779