redo and next should raise a SyntaxError in methods

[eregon]

See dcb41d0 and the 2 previous commits.

Should it be done in the parser? (and is it reasonable?)
It seems MRI does it slightly later as -c or Ripper will not catch it for instance.

[enebo]

@eregon We do not have parent links between node in AST so I think this is more easily done after parsing but I don't want the parser to walk the tree after making it to look for this condition (it would slow startup to re-walk each tree). This means both Truffle and IR should detect this during the assembling their respective intermediate forms. This is a little duplication but can be done when naturally walking the tree to build our IRs.

[enebo]

@eregon I guess we could do a subtree visit on DefnNode creation but it still feels like an extra traversal of part of the tree.

[eregon]

@enebo Sounds reasonable, I actually implemented it in Truffle translator.
I was wondering whether the parser has the knowledge if it is within a block/method or not. I agree an extra traversal is not worth for this, it can be dealt with later easily.

[enebo]

@eregon At the time we are creating the node we have no up-pointing references to higher scope types (e.g. begin, def{sn}, module,class, etc...) so we need to walk from scopes down to look for the redo/next. We create staticScopes but they have no back references so that is not usable either.