Bugfix: Rough edges in JFactoryPodioT and JMultifactory

[nathanwbrei]

• JMultifactory output collection names are no longer suffixed with mTag. We are now handling this at the JOmnifactory level instead. Technically this would be a breaking change, but we never used this feature to begin with.
• JFactoryGenerator is now given its own mApp pointer, so that we no longer need to pass it in to all of the J{Omni, Chain}FactoryGenerators.
• JFactoryPodioT::Set(std::vector<T*>) and Insert(T*) no longer leak memory
• JMultifactory correctly catches std::exception and wraps it in a JException, giving us better stack traces
• PODIO performance and unit tests now support both old and new PODIO versions
• JMultifactory was verified to have a usable mPluginName and mApp
• JFactoryPodioT::Insert() was verified to work correctly
• A compiler warning involving C++20 implicit capture of this has been fixed

This PR addresses almost everything in the following issues:

• Small fixes for JOmnifactory #263
• Remaining PODIO integration problems #247
• Memory leak in JFactoryPodioT::Set #235

The only remaining issue is in #247, regarding Set<>() not checking if the collection is already in the frame. This issue is pretty subtle: a user would have to go reasonably far off the beaten path for it to crop up, and it's not obvious to me how JANA ought to recover, anyhow. I think we can worry about it after this release. It would make the most sense to revisit the issue when we address issue #254.