Fix early exit with stop-at-end when commits remain in queue

[enmaku]

Summary

When rendering to video (-o), Gource can stop the simulation before the final commits are processed if there is a long idle gap in the repository history.

This changes two guards in Gource::logic() / Gource::updateUsers() so that stop_at_end still waits for commitqueue to drain, and auto-skip remains enabled while queued commits are waiting.

Problem

-o (and --stop-at-end) enable stop_at_end unless --dont-stop, --loop, or reading from stdin (-) is used (gource_shell.cpp).

In readLog(), once the commit log reaches EOF (commitlog->isFinished()), stop_position_reached is set immediately — even though commits with later timestamps may already be sitting in commitqueue waiting for simulation time to catch up.

Two consequences:

1. Auto-skip is disabled once stop_position_reached is true (logic(), ~line 1749), so Gource no longer jumps across idle gaps while those queued commits are pending.
2. Exit does not check the queue — updateUsers() sets appFinished when users.empty() && stop_position_reached (~line 1401), without requiring commitqueue.empty().

On a repo with a multi-day gap followed by a burst of commits (e.g. last activity on 24 June, next commit on 30 June), the visualization can exit around the gap while the final commits never appear. The on-screen clock reflects the last processed time, not the last commit in the log.

Reproduced with stock Gource 0.56 on a real git repo: gource --start-date '2026-06-20' --stop-date '2026-07-01' -1280x720 -o - <repo> — last frame showed ~26 June; the 30 June commits were never rendered.

Fix

1. Exit guard — only finish when the queue is drained:

   if(users.empty() && stop_position_reached && commitqueue.empty())

2. Auto-skip guard — keep skipping idle time while commits remain queued after EOF:

   if(... && (!stop_position_reached || !commitqueue.empty()))

No changes to log parsing, timestamp handling, or CLI flags.

Test plan

• Repo with a multi-day commit gap and activity after the gap: render with -o and confirm the final commits and clock date are shown
• Same repo without -o / --stop-at-end — behaviour unchanged
• --dont-stop and --loop paths unchanged
• Short history with no gaps — still exits cleanly at end
• --stop-date / --stop-position stopping before EOF — still stops at the requested point

Verified locally on macOS (arm64) against a git repo with a ~5.7-day gap (24 Jun → 30 Jun): stock build stopped ~26 Jun; patched build reached 30 Jun and exited on its own.