This PR does a few things:

• Excludes endpoints whose path has examples in it. This removes a lot of example code that otherwise gets picked up when modeling libraries.
• Changes the FrameworkModeEndpoints query to produce not only methods, but also classes. These are not technically endpoints, but it allows the model editor to learn about them and thus handle the new results from GenerateModel (see below). This involves some refactoring of the Endpoint class, because we want to include both DataFlow::MethodNodes and DataFlow::ModuleNodes, which don't have a common ancestor.
• Updates the GenerateModel query to produce typeModel rows for subclass relationships. For example:

class A
  def foo(x)
    # ...
  end
end

class B < A
end

If we mark A#foo as a sink, we also want to recognise calls to B#foo as a sink as well. Instead of duplicating the models for every subclass, we instead emit a type row:

A,B,

which states that B can be considered to have type A. This means any models that target A will target B as well.

The same applies for include:

module C
  def bar(x)
    # ...
  end
end

class D
  include C
end

but not for extend, because its behaviour is a bit odd (instance methods on the included module become class methods on the including class). I haven't worked out how to deal with them yet.