Main goals of the whole model refactoring:
- For NifItem and *Model (BaseModel, NifModel, etc.) methods with a QString parameter provide QLatin1String and 'const char *' overloads. QLatin1String("foo") is much faster than "foo" with its under-the-hood conversion to QString because in Qt 5 QLatin1String("foo") could be done at compile time. And this will make something like reading Skyrim SE shapes several (4 to 8) times faster.
- For *Model methods working with QModelIndex child items provide convenient overloads to work with 'NifItem *' instead. Overall, gradually migrate from using QModelIndex to 'NifItem *', at least on low level, to avoid back-and-forth conversions between QModelIndex and 'NifItem *' and to simplify the code overall.

Bonuses in this commit:
- NifItem: name(), type(), ... getters now return 'const QString &' instead of QString, to avoid wasting resources on creating a QString copy when it's not needed.
- NifItem: new hasName and hasType methods as a replacement for 'name() == "foo"' and 'type() == "foo"' ('const char *' overloads - for performance (see main goal #1), QString and QLatin1String - for uniformity).
- NifItem: new get<T> method as a shortcut for this->itemData.value.get<T>().
- NifItem: static versions of get<T> and getArray<T> methods, with a nullptr check for the item.