Grammar changes

• Introduce TypeExpression, a node type that wraps a Type node but is an ASTExpression

  • This allows us to treat instanceof expressions as infix expressions (which they are, syntactically), and fix the weird inconsistency, that instanceof is in the middle of the precedence hierarchy of infix operators, yet is not itself an ASTInfixExpression. Now it has its own constant in BinaryOp, and ASTInstanceOfExpression extends ASTInfixExpression (its XPath name is InfixExpression too).
  • This also simplifies the qualifier of method call, method references, and field accesses. Previously it could have been a type, an expression, or an AmbiguousName (which implements both interfaces). Now we can use the single interface ASTExpression to unify them, and merge the methods getLhsType, getLhsExpression, getAmbiguousLhs (which were inconsistently defined anyway).
  • This is what Spoon does

• Cleanup inconsistencies in the API of many nodes

  • align the API of ASTAssignmentExpression on that of ASTInfixExpression
  • rename the methods of ASTConditionalExpression to a simpler scheme (I'd like ASTIfStatement to have the same names)
  • Fix many doc comments or simplify them
    • In particular, fix many comments specifying that some node is a PrimaryExpression, or a node with a specific precedence. Since parentheses are flattened, any expression may appear there (even those that have a lower precedence, provided they were parenthesized). There were quite a few bugs where eg getLhsExpression expects an ASTPrimaryExpression even though anything can be there.

TBH ASTPrimaryExpression is not useful in the API since #1872, and we should probably scrap it entirely.