Add error handling for run-level profiler initialization

[Copilot]

Description

Addresses feedback from regarding missing error handling when initializing the run-level profiler via RunOptions.

Changes:

• Wrapped profiler initialization in try-catch to handle exceptions from Initialize() and StartProfiling()
• Added validation after StartProfiling() to verify IsEnabled() returns true
• Reset run_profiler and log warnings on failure, allowing execution to continue without profiling

Before:

if (run_options.enable_profiling && !session_profiler_.IsEnabled()) {
  run_profiler.emplace();
  run_profiler->Initialize(session_logger_);  // Can throw
  // ... setup code ...
  run_profiler->StartProfiling(profile_file);  // Can fail silently
}
// Later: assumes run_profiler is valid if set

After:

if (run_options.enable_profiling && !session_profiler_.IsEnabled()) {
  try {
    run_profiler.emplace();
    run_profiler->Initialize(session_logger_);
    // ... setup code ...
    run_profiler->StartProfiling(profile_file);
    
    if (!run_profiler->IsEnabled()) {
      LOGS(*session_logger_, WARNING) << "Failed to enable run-level profiler...";
      run_profiler.reset();
    }
  } catch (const std::exception& ex) {
    LOGS(*session_logger_, WARNING) << "Failed to initialize: " << ex.what();
    run_profiler.reset();
  }
}

Motivation and Context

Without error handling, profiler initialization failures (e.g., null logger, file open failures) would result in undefined behavior when dereferencing run_profiler later in the execution path. The fix ensures graceful degradation when profiling cannot be enabled.

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.