fix crash on corrupted transition name

[summeroff]

Description

Net change: addSource for transitions now owns ti via unique_ptr from the start, calls try_emplace before connecting any signal handlers, and connects only if insertion actually happened — using result.first->second.get() (the map-owned pointer) as the user-data. On duplicate, the local unique_ptr destructs cleanly, no handlers ever reference a freed TransitionInfo, and a warning is logged so the situation is visible in the next dump.
Also adds a NULL-guard on obs_source_get_name — kills the std::string(nullptr) UB that was a third trigger for the same crash path.

Motivation and Context

How Has This Been Tested?

Types of changes

• Bug fix (non-breaking change which fixes an issue)

Checklist:

• The code has been tested.
• All commit messages are properly formatted and commits squashed where appropriate.
• I have included updates to all appropriate documentation.

[Copilot]

Pull request overview

This PR hardens transition tracking in CallbackManager::addSource to prevent crashes when transition names are corrupted/duplicated, by ensuring map ownership is established before signal connections and by avoiding std::string(nullptr) construction.

Changes:

• Adds a null-guard around obs_source_get_name() for transitions before logging and key construction.
• Reworks transition insertion to use std::unique_ptr + try_emplace, connecting signal handlers only when insertion succeeds and using the map-owned pointer as callback userdata.
• Logs a warning when a transition with the same name is already tracked.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.