[bettodiaz]

Summary

A Flow handoff reports TIMEOUT and skips cursor placement when the switched devices take longer than 3 s to show up on the receiving side, even though they do arrive. I measured how long that takes:

• Linux receiver: 2.0 to 4.2 s per handoff, so roughly half of them missed the 3 s window.
• macOS receiver: 7 to 9 s (the Mac agent log shows its 3 s window closing, then the devices appearing about 4 to 6 s later).

This raises the receiver's arrival window from 3 s to 10 s, the existing MAX_ARM_TIMEOUT a sender honors.

Changes

• openlogi-agent-core: ARM_TIMEOUT 3 s to 10 s, with a comment giving the measured arrival times; the one handoff test that hard-coded the 3 s window now follows the constant.

Testing

• cargo fmt --all -- --check, cargo test -p openlogi-agent-core -p openlogi-agent (1065 passed): pass.
• cargo clippy -p openlogi-agent-core -p openlogi-agent --all-targets -- -D warnings: fails on 8 map_unwrap_or lints in openlogi-core that are already on master (Rust 1.100); clean with that one lint allowed.
• Not run: the full workspace gate, ast-grep, typos, cargo-deny, the macOS and Windows jobs.
• Hardware: with an 8 s window on a Linux receiver, all six returns from a Mac arrived (2.0 to 4.2 s) and the cursor was placed each time. The Mac side was not tested with the new value.