Hi, and thanks a lot for the amazing work on this PHP parser!

I am writing a generic tree transformer for tree-sitter grammars in order to integrate them in a structural diff tool (https://github.com/GumTreeDiff/gumtree) and I was currently trying to integrate PHP thanks to your awesome grammar.

I noticed something weird for the case of strings. If I parse

<?php foo("test") ?>

I obtain (syntax is this: type: associated_text [start_pos,end_pos], I only retain the associated text on leafs)

program [0,21]
    php_tag: <?php [0,5]
    expression_statement [6,17]
        function_call_expression [6,17]
            name: foo [6,9]
            arguments [9,17]
                (: ( [9,10]
                argument [10,16]
                    encapsed_string [10,16]
                        ": " [10,11]
                        string: test [11,15]
                        ": " [15,16]
                ): ) [16,17]
    text_interpolation [18,20]
        ?>: ?> [18,20]

Which is perfect! (perhaps thinking again I am not sure to understand the text_interpolation node but this is not a problem for my usage)

However for the program

<?php foo('test') ?>

I get

    php_tag: <?php [0,5]
    expression_statement [6,17]
        function_call_expression [6,17]
            name: foo [6,9]
            arguments [9,17]
                (: ( [9,10]
                argument [10,16]
                    string [10,16]
                        ': ' [10,11]
                        ': ' [15,16]
                ): ) [16,17]
    text_interpolation [18,20]
        ?>: ?> [18,20]

Notice that in this case the value associated to the string production is not in any child of the string typed node, as it was the case for the encapsed_string and it prevents me to associate the string value to the string node since I retain only the text for leaf nodes.

Is it a normal behavior? Would it be too much hassle to make it work like encapsed_string work?

Cheers!