use <wchar.h> with C++ linkage on Windows ARM

Instead of commenting that we should, just do include <wchar.h>
with C++ linkage when compiling for Windows on ARM.
Actually, omitting the C linkage really should be enough here.

Fix #483 by using the correct value type

fix typos and dead links in docs

GenericDocument: add implicit conversion to ParseResult

To simplify the error handling, this commit adds an implicit conversion
of GenericDocument to ParseResult, allowing code like (already in the
documentation):

   ParseResult ok = doc.Parse(json);
   if (!ok) // ...

fix Document::Parse(const Ch*) for transcoding

To allow for an UTF16-Document to Parse(UTF8),
the Parse() argument has to be UTF8-compatible.

documenttest.cpp: EXPECT_THROW when checking empty allocator (closes #469)

In the MoveConstructor/MoveAssignment tests, a check for a `NULL` allocator
return has been used to check for the correct moved-from state of a Document.
After the merge of #426, the GetAllocator call fails with an assertion,
if the allocator of a GenericDocument is NULL.

Check for the throw, instead of NULL in the move-related tests.
Tested with GCC 4.9 on Linux with C++11 enabled.

Reported-by: @woldendans

Adding coapp definition

Adding usage comment at the top of the file for knowledge sharing

https://github.com/miloyip/rapidjson/pull/460#issuecomment-152232015

To create a nupkg, call CoApp's Write-NuGetPackage .\rapidjson.autopkg
-defines:MYVERSION=1.0.2 (or whatever version you want to call it)

Highlight main features in bold

Fixed typo

Add missing return statement, fix #448

Add break at default switch case statements fix #444

fix issue #444

Introduce comments support

* Comments parsing function correctly handles EOF.
* Since SkipWhitespaceAndComments can generate errors, its calls should be followed by RAPIDJSON_PARSE_ERROR_EARLY_RETURN macro.
* Some tests to make the bug never appear again.

<algorith> should be <algorithm>.

Might help some people see things better when skimming.

Keep Document value unchanged on parse error, fixes #437

Keeping the DOM unchanged in case of an error is the intended
behaviour according to the [documentation] [1].

Instead of forcing the value to `kNullType` before starting the
parsing, store the parsed value upon success via regular move.

[1]: https://miloyip.github.io/rapidjson/md_doc_dom.html#ParseError

Fix UB in GenericDocument::ParseStream

To check is it possible to expose allocator.