unbust extra scope being introduced

When using the sugared object property form `{ say }`, the
property:method rule triggered (not the property:identifier rule
-- I think the former won simply by dint of being first), and
instead of failing outright, it got as far as entering a new
lexical scope, and *then* failing.

Which means that the function declared in the test code got
declared in a scope that was further nested than the one in
which it got called:

    f();
    my o = { say };
    sub f() { say("Mr. Bond") }

And so it failed.

I *think* the lesson for future grammar authors is this: don't
run any side effects until you're sure you're in the right
rule. In this case, `property:method` had time to run a side
effect, and then immediately fail before the backtracking
carried the parser into the right rule, `property:identifier`.

Closes #150.

lib/_007/Parser/Syntax.pm

@@ -250,9 +250,11 @@ grammar _007::Parser::Syntax {
     rule property:identifier-expr { <identifier> ':' <value=EXPR> }
     rule property:method {
         <identifier>
-        :my $*insub = True;
-        <.newpad>
-        '(' ~ ')' <parameterlist>
+        '(' ~ ')' [
+            :my $*insub = True;
+            <.newpad>
+            <parameterlist>
+        ]
         <trait> *
         <blockoid>:!s
         <.finishpad>

t/features/objects.t

@@ -239,4 +239,17 @@ use _007::Test;
         "`new` on object literals without the type is allowed but optional";
 }
 
+{
+    my $program = q:to/./;
+        f();
+        my o = { say };
+        sub f() { say("Mr. Bond") }
+        .
+
+    outputs
+        $program,
+        qq[Mr. Bond\n],
+        "using the short-form property syntax doesn't accidentally introduce a scope (#150)";
+}
+
 done-testing;