Fixes some formatting for decorators

[rbuckton]

Fixes #2660 and #2661

• Added rules for spacing around decorators
• Computing the undecorated start line of a node for use as the effectiveParentStartLine

There's still an issue with indentation in formatOnEnter.

[mhegazy]

i made a change to make isExpression return true for decorators.

[mhegazy]

can you just walk up until you find a decorator, or just use hasAncestor/getAncestor

[mhegazy]

here is the change i referenced earlier: #2659

[rbuckton]

Based on our offline discussion, I changed the logic in isExpression in #2659. This should no longer be an issue.

[rbuckton]

If we ever add support for decorators on class expressions, function expressions, or arrow functions then hasAncestor/getAncestor is unsafe to use. Consider the following:

@class { /*A*/ } class C { }

At A above, a call to getAncestor(SyntaxKind.Decorator) would return the decorator for C, but it should have stopped at the class body. hasAncestor and getAncestor don't have a boundary. We could add a boundary check, but this is easy enough.

I may change this to the following, though:

        static NodeIsInDecoratorContext(node: Node): boolean {
            while (isExpression(node)) {
                node = node.parent;
            }
            return node.kind === SyntaxKind.Decorator;
        }

[mhegazy]

👍