Protect event loop from user callback exceptions

[VolkerChristian]

Motivation

• Prevent uncaught exceptions in user callbacks from unwinding the event loop and crashing the process.
• Avoid leaving composite resources (e.g. socket reader/writer pairs) in half-open states when one side throws.
• Provide a central recovery hook so individual receiver types can implement domain-appropriate teardown behavior.

Description

• Add a global safety net in core::Event::dispatch that catches all exceptions from onEvent and forwards them to a new recovery hook: EventReceiver::onEventException(std::exception_ptr). (files: src/core/Event.cpp, src/core/EventReceiver.h)
• Implement a default EventReceiver::onEventException that logs the exception (standard and non-standard) using the logging facility. (file: src/core/EventReceiver.cpp)
• For descriptor-backed receivers, override the recovery hook so a failing callback causes the receiver to be disabled, preventing repeated failing callbacks on the same fd. (files: src/core/DescriptorEventReceiver.h, src/core/DescriptorEventReceiver.cpp)
• In socket stream code, override the hook for the SocketReader/SocketWriter paths to route exceptions into the existing error handling path by invoking onStatus(EIO), which triggers the established connection teardown flow and avoids leaving only one side disabled. (files: src/core/socket/stream/SocketReader.{h,cpp}, src/core/socket/stream/SocketWriter.{h,cpp})

Testing

• Ran git diff --check to verify no whitespace or diff errors, which passed.
• Ran cmake -S . -B build to validate project configuration; configuration failed in this environment due to a missing external dependency (nlohmann_json>=3.11) unrelated to these code changes, so a full build was not completed here.

---

Codex Task