Multi-Threading is Broken - Needs Simplification

[EthanMBoos]

What's wrong

1. The main bug: The worker thread uses condition_variable::wait() without a predicate. This means threads can wake up spuriously, see an empty queue, and bail out - silently dropping tasks. Entities randomly don't get stepped.

2. No RAII locks: There are 36 manual lock()/unlock() pairs. If anything throws an exception, the mutex stays locked forever → deadlock.

3. Too many mutexes: 10 separate mutexes protecting simple boolean flags and related timing data. This is overkill and creates complex ordering rules.

The Fix

Convert 5 mutexes → atomics (for simple bool flags)

Delete This Mutex	Replace With
finished_mutex_	std::atomic<bool> finished_
exit_mutex_	std::atomic<bool> exit_
paused_mutex_	std::atomic<bool> paused_
single_step_mutex_	std::atomic<bool> single_step_
take_step_mutex_	std::atomic<bool> take_step_

Consolidate 3 timing mutexes → 1

These all protect related timing state and are often needed together (computing actual time warp requires both elapsed wall time and sim time atomically). Also, time_warp_mutex_ is declared but never used.

Delete These	Replace With
timer_mutex_	timing_mutex_ (single mutex)
time_mutex_
time_warp_mutex_

Keep 2 mutexes (with improvements)

Mutex	Change
contacts_mutex_	Convert to std::shared_mutex (concurrent reads)
entity_pool_mutex_	Keep as-is, fix the condition variable predicate

Other changes

• Replace all 36 manual lock()/unlock() calls with RAII (lock_guard/unique_lock)
• Add predicate lambda to condition_variable::wait() to fix spurious wakeups

Total: 10 mutexes → 3 mutexes + 5 atomics