Atomics, improve lisa.fol.FOL's logic and simplify it

[SimonGuilloud]

The PR does a few things:

• Rename old and poor traits and classes in the kernel's FOL and front's FOL, following the realization the the word I was looking for is exactly "Atomics". For example, "SchematicVarOrPredLabel" --> "SchematicAtomicLabel". This does not change any functionality.
• Fixed the logic in the front's FOL. Labels were contravariant in their type parameters when they should not be; This was necessary for fixing some classes extending labels twice, itself following some other workarround. This as all been fixed. Moreover, apply paterns on Variables, Constants, Terms and Formulas should be much cleaner and logical. The cost was to carry a little bit more type parameters, but it's well worth it.
• Removed the match type for T *** N, which was used to obtain function application with arbitrary but fixed number of parameters given by N. Instead, T**N is now always a sequence at runtime, hidden behind an opaque type. Function application with N arguments is obtained through extension methods (one for each N, not one for each implementation of (T ** N) |-> S). That's a bit less general, but more flexible and possibly more efficient.
• Generally refactored the Common.scala file, ensuring uniformity across definitions. Uses less top-notch, scala-specific features and should be a bit more stable and easier to understand.

[sankalpgambhir]

Looks good.

Just out of curiosity, do you notice a significant change in compile times with a move to hard coded arities over the previous generic framework?