ostest: reduce time and fix bugs cause by reducing time

[zzby0]

Note: Please adhere to Contributing Guidelines.

Summary

This PR optimizes the ostest suite execution time and fixes several reliability issues:

1. Reduce test execution time: Reduced sleep() durations across multiple test cases, decreasing total test time from ~150s to ~60s (60% reduction) while maintaining test effectiveness.

2. Fix timing-related test failures: Replaced timing-based synchronization (usleep/sleep) with deterministic event-based synchronization (semaphore polling, waitpid) to eliminate race conditions on slower systems or under high load.

3. Fix SMP-related race conditions:

   • Fixed robust_test failure in SMP environments where parent thread could acquire mutex before child thread
   • Fixed restart_test assertion in semaphore code by restructuring the test to ensure restart happens under controlled conditions

Changes by category:

• Time optimization: 24 files with reduced sleep/delay times
• Synchronization improvements: sighand.c, suspend.c
• Race condition fixes: robust.c, restart.c
• Timeout adjustments: pthread_rwlock.c

Impact

Users:

• Faster CI/test execution (60% time reduction)
• More reliable test results across different hardware speeds and configurations
• Better test coverage for SMP systems

Build process: No impact - pure test code changes

Compatibility: Fully backward compatible - no API or behavior changes to tested components

Testing reliability: Significantly improved - eliminates timing-dependent race conditions that caused intermittent failures

Testing

Verification performed:

• All modified test cases executed successfully on both single-core and SMP configurations
• Tests verified under various load conditions to ensure timing changes don't cause false failures
• Confirmed 60% reduction in total ostest execution time (150s → 60s)

Test configurations verified:

• Single-core ARM Cortex-M systems
• Multi-core SMP configurations
• Simulator (sim:ostest)

Before changes:

• Test time: ~150 seconds
• Intermittent failures observed in:
  • robust_test on SMP systems (timing race)
  • sighand_test on slower systems (fixed delay insufficient)
  • restart_test (assertion in semaphore code)

After changes:

• Test time: ~60 seconds (60% improvement)
• All tests pass reliably across configurations
• No false positives due to timing issues
• SMP race conditions eliminated

Build verification:

• Host: Linux x86_64
• Target: Multiple architectures (ARM, RISC-V, simulator)
• All builds successful with no new warnings

The changes maintain test coverage while significantly improving execution speed and reliability.