[ian-at-strix]

What

Re-lands #80, #81, #89 and #75, one commit each, in that order.

Why

#80, #81, #89 and #75 merged into stacked branches after their bases had landed, so main never got them.

How

Cherry-picked each PR's own commit from windows/logo onto main, kept main's code where #68's reformat conflicted, then cargo fmt. #75 is the version stacked on #81's task dialog (what merged), not the standalone rebuild. Dependency changes redone with cargo add; Cargo.lock matches the original.

Tested

• macOS: cargo build, cargo clippy --all-targets, cargo test, cargo fmt --check, with and without machine-wide
• cargo clippy --target x86_64-pc-windows-msvc on tray.rs and pipe.rs in a scratch crate with main's windows features
• dist plan

Untested

• A Windows build (build.rs embedding the manifest and icon) and the tray, popups and UAC prompt on Windows