As reported in #113, recent versions of Clang complain about
ambiguous overloads for some comparison operator instantiations,
especially if the deduced template type is a GenericValue.

Add an explicit, non-templated version for now (which is a better
match).

This only solves part of the problem, as comparisons between
  * GenericValue & GenericDocument
  * GenericValue with different SourceAllocator types
will still cause ambiguities.

GenericMemberIterator: fix comparisons/differences for mixed-constness expressions

Parsing optimization

Do unaligned non-SIMD matching first, and then SIMD matching later.
Also add a fast path for no skipping.

GenericValue: add optional support for std::string

Some users may want to use RapidJSON with std::string objects.  This commits
adds an (opt-in) feature to include some basic support.  The implementation
uses std::basic_string<Ch> as generic string type.

Support currently covers:
 * construction
 * comparison

No special APIs for AddMember or PushBack have been added, as std::string
most probably requires copying (or an explicit StringRef() call).

By restructuring the call forwarding of the various operator== and
operator!= overloads, new overloads can be added by simply adding an
additional member operator==.

Additionally, the "Ch*" overloads are dropped in favour of an SFINAE
version that removes the pointer variants from matching the templated
operator== (see also operator=).

Add Doxygen documentation for configuration macros and error handling

Memory optimization

Use CrtAllocator for stack.
ShrinkToFit stack after parsing.

Lazy allocation. (so Document not for parsing don't need to allocate
stack).
Remove redundant stackCapacity member variable.

.travis.yml: enable 32-bit builds

Use a tighter bound.

As Ubuntu supports running 32-bit executables based on its multiarch
setup, we can enable the tests on 32-bit configurations as well.

Currently, valgrind fails for 32-bit executables on the 64-bit host.
This requires some more investigation.

Move to RFC7159 (closes #90)

Fix sign-comparison warning/error on 32-bit platforms

By adding an optional CRTP template parameter, the BaseReaderHandler
can call the "overridden" `Default()` function from the `Derived`
class.

See https://en.wikipedia.org/wiki/Curiously_recurring_template_pattern

With the allowance of arbitrary root value types, the individual tests
can use the top-level Parse functions, instead of ParseFoo() variants.

Secondly, some unneeded array wrappers have been dropped and non-singular
tests starting with other values than objects or arrays have been added.