feat: use systemd-journal-logger when running as systemd service

[erictapen]

This produces nicer logging output with colors for log levels and no duplicate time stamps.

Currently it looks like this:

Aug 21 14:03:34 maschine systemd[2827]: Started Passless FIDO2 Software Authenticator.
Aug 21 14:03:34 maschine passless[151280]: [2026-08-21T12:03:34.633Z INFO  passless_core::config] Loading configuration from: /nix/store/xgfdwsd064n55ycjn8c1vamicr73qyn7-passless.toml
Aug 21 14:03:34 maschine passless[151280]: [2026-08-21T12:03:34.633Z INFO  passless] Enabling verbose logging...
Aug 21 14:03:34 maschine passless[151280]: [2026-08-21T12:03:34.633Z DEBUG passless] Verbose logging enabled
Aug 21 14:03:34 maschine passless[151280]: [2026-08-21T12:03:34.633Z INFO  passless] Applying security hardening...
Aug 21 14:03:34 maschine passless[151280]: [2026-08-21T12:03:34.633Z DEBUG passless_core::config] Disabling core dumps to prevent credential leakage
Aug 21 14:03:34 maschine passless[151280]: [2026-08-21T12:03:34.633Z DEBUG passless_core::config] Check mlock capability
Aug 21 14:03:34 maschine passless[151280]: [2026-08-21T12:03:34.633Z DEBUG passless_core::config] MLOCK is enabled - sensitive data will not be swapped to disk
Aug 21 14:03:34 maschine passless[151280]: [2026-08-21T12:03:34.634Z INFO  passless] Acquiring instance lock...
Aug 21 14:03:34 maschine passless[151280]: [2026-08-21T12:03:34.634Z DEBUG passless] Instance lock acquired at /run/user/1001/passless/1acd256aa9f526d1cd5a3aab21ccd746.lock
Aug 21 14:03:34 maschine passless[151280]: [2026-08-21T12:03:34.634Z INFO  passless] Creating UHID device...
Aug 21 14:03:34 maschine passless[151280]: [2026-08-21T12:03:34.734Z INFO  passless] Creating authenticator service...
Aug 21 14:03:34 maschine passless[151280]: [2026-08-21T12:03:34.734Z INFO  passless::storage::pass] Using pass (password-store) backend
Aug 21 14:03:34 maschine passless[151280]: [2026-08-21T12:03:34.734Z INFO  passless::storage::pass] Store path: /home/kerstin/.local/share/password-store
Aug 21 14:03:34 maschine passless[151280]: [2026-08-21T12:03:34.734Z INFO  passless::storage::pass] Path: fido2
Aug 21 14:03:34 maschine passless[151280]: [2026-08-21T12:03:34.734Z INFO  passless::storage::pass] GPG backend: gpg

[pando85]

Thanks for the PR, this will be an improvement in our logging stack.

[pando85]

Thanks for this — I like the direction and the native journal integration is a real improvement. I’d like to get this merged; there are just a few small compatibility details I’d like us to keep from the current logging behavior:

1. Preserve PASSLESS_LOG_LEVEL when journald is active. Right now the journal branch bypasses the env_logger filter entirely, so values such as warn or module-specific filters stop working under systemd. We can keep the native JournalLog and use an env_logger::Logger built with Builder::build() purely as the filtering layer in front of it (it implements log::Log and can be nested), so no extra dependency should be necessary.

2. Fall back instead of panicking if the native journal socket is unavailable. connected_to_journal() only checks JOURNAL_STREAM against stderr; JournalLog::new() separately connects to the native journal socket and can fail, for example in some container/root-image setups. If JournalLog::new() fails, falling back to the existing stderr env_logger path would keep Passless usable. A short eprintln! explaining the fallback would be useful too.

3. Keep the agent journal identifier. contrib/systemd/passless-agent.service intentionally uses SyslogIdentifier=passless-agent, while the new native logger always writes SYSLOG_IDENTIFIER=passless. Because native journal fields bypass systemd’s stdout/stderr SyslogIdentifier= handling, agent logs would become passless. I suggest an env override such as PASSLESS_SYSLOG_IDENTIFIER (defaulting to passless) and setting Environment=PASSLESS_SYSLOG_IDENTIFIER=passless-agent in the agent unit. The normal service can keep the default.

PASSLESS_LOG_STYLE can remain relevant only to the stderr fallback; native journald already carries the priority structurally, so that part does not need to be emulated.

There is also an unrelated getrandom lockfile bump through tempfile; nice to avoid if Cargo lets us, but I would not block the PR on that.

Thanks again for the contribution — the core approach is good; these are mostly Passless-specific integration details rather than a change in direction.

[erictapen]

1. Shouldn't log::set_max_level already do this?
2. done
3. Couldn't we just detect wether we are running as root and set the syslog identifier accordingly?

Feel free to squash this when merging.