Describir: Exploring Attention Patterns and Neural Activations in Transformer Architectures for Sequence Classification in Context Free Grammars