cuprated: Add graceful shutdown, pt3: Exit codes and main error propagation

[redsh4de]

What

Depends on #586

Building on part 2, this replaces hard exit and panics across cuprated's startup path with error propagation, introduces ExitCode to distinguish clean shutdowns from failures.

Based off #528
Closes #522.

Where

• cuprated
  • main.rs - main() returns ExitCode, new main_inner() propagates all startup errors via ?, removed all process::exit(1) calls from config/init path
  • monitor.rs -TaskExecutor now has failed: Arc<OnceLock<&'static str>>, spawn_critical records the failing task on failure, typed CriticalTaskError for exit code decision
  • lib.rs - Node::wait_for_shutdown returns Result<(), CriticalTaskError>; returns CriticalTaskError if a critical task failed after the drain

How

main_inner() returns anyhow::Result<ExitCode>. Startup errors propagate to main(), which prints them with eprintln_red and returns ExitCode::FAILURE.

For runtime errors, spawn_critical records the failing task's name in a OnceLock on the TaskExecutor before triggering shutdown.wait_for_shutdown checks it after draining all tasks and returns Err(CriticalTaskError) if set - that Err flows through ? in main_inner to main(), which prints and returns ExitCode::FAILURE. Clean shutdowns return ExitCode::SUCCESS.