Fix bug #70570: json_encode and var_dump ignore simplexml cdata

[nielsdos]

A sanity check for the behaviour would be nice :)
Sending to master because of potential BC break.

XML_CDATA_SECTION_NODE was simply not handled, handle it the same way as we handle text nodes for consistency reasons.
We introduce the php_sxe_is_inclusive_text_node() helper for text-like nodes. In DOM parlance, these 2 types are called inclusive text nodes. Unfortunately, the fact that we handle CData the same as text now has a technical BC break.

Previously this XML:

<?xml version="1.0" encoding="UTF-8"?>
<container>
    <C><![CDATA[hello]]><foo/><![CDATA[world]]></C>
</container>

resulted in this var_dump output:

object(SimpleXMLElement)#1 (1) {
  ["C"]=>
  object(SimpleXMLElement)#2 (1) {
    ["foo"]=>
    object(SimpleXMLElement)#3 (0) {
    }
  }
}

However, after this patch (as text is not an element and thus handled specially) we get the following output:

object(SimpleXMLElement)#1 (1) {
  ["C"]=>
  string(10) "helloworld"
}

[nielsdos]

Converting to draft, as name may not be set and therefore getName() may do a null access. (I think)

[Girgias]

I find it strange that text nodes are not displayed in the first place. I wonder what the rationale for that is.

[nielsdos]

A general solution for the inconsistencies in simplexml would be better, and would likely need to go through an RFC because it can have BC impact.