Do not write out invalid characters to XML file.

[rodrigc]

For special characters such as: <, >, or ', write them out
to the file as:
    &#[decimal ASCII value];

Character     Written to    Appears in Jenkins
to encode     XML file      test result viewer
=========     ==========    ==================
  <            &lt;             <
  >            &gt;             >
  '            &apos;           '
  &            &amp;            &
  "            &quot;           "

For characters which are listed as RestrictedChar
in http://www.w3.org/TR/xml11/#charsets , these characters
are completely invalid XML, and cannot even be escaped.

Character     Written to    Appears in Jenkins
to encode     XML file      test result viewer
=========     ==========    ==================
  0x08         &amp;#8;          &#8;
  0x1F         &amp;#31;         &#31;

This will at least generate a valid XML file,
but let us see where these restricted characters would
appear in the output.

Fixes #136

[jmmv]

Why bother respecting all these entities? Why not just use the decimal ASCII value in all cases? The latter would keep the code simpler, and the resulting XML wouldn't depend on the definition of the entities, which are not available in a "plain" XML document.

[rodrigc]

I respected those entities, because the old code did that.
Changing this behavior to emit ASCII for these characters would be a change
in behavior, and would require updating the utils/text/operations_test:escape_xml__some_escaping unit test.
I didn't want to go down that path, but if you want that, and allow me to change the unit test, I can do that.

[jmmv]

OK, I guess it makes sense to leave as you did. I read the spec links you provided and it seems that there is a clear difference between the "special characters" and everything else.

I personally like the named entities better for readability, but the introduction of two behaviors (named entities for these, numeric for the rest) is a bit unfortunate.

[jmmv]

This change is missing tests. The existing tests for escape_xml need to be extended to ensure that a variety of non-alphanumeric characters are properly handled.

And super-nitpick: the change description does not follow the guidelines in CONTRIBUTING.md:

• Follow standard Git commit message guidelines. The first line has a maximum length of 50 characters, does not terminate in a period, and has to summarize the whole commit. [...]

[jmmv]

The cast would better be static_cast< std::string::size_type >(*it) to not make any assumptions on the character/integer sizes.

[rodrigc]

I added that.

[jmmv]

What's the point of the double "" in the middle of the argument passed to escape_xml? Seems unnecessary.

[rodrigc]

It's needed because otherwise the compiler will think that it is \x7fBA

[jmmv]

Committed.

I did a manual merge to add a proper entry to NEWS as part of your commit and to drop trailing whitespace from the code (which GitHub helpfully did not show, but I saw on the command line).