This is largely motivated by the fact that Clang currently miscompiles `.then([](auto){})`:

https://llvm.org/bugs/show_bug.cgi?id=24989

As of that change, adopting an empty struct into a pointer that already had some other value fails to overwrite the upper 32 bits of the pointer, thus resulting in a struct pointer with totally bogus bounds. Normally this will either cause a bounds check exception to be thrown when that pointer is later accessed, or lead to a struct containing bogus data (but if teh struct was expected to be empty, probably that data is never accessed).

The MSVC runtime was heavily refactored, so the internal function we call to
get the current exception count changed for newer MSVC versions (as did the
offset).

This encapsulates the input array, hiding it from the parser's methods.
Access is forced to go through `advance` and `consume*` methods on the
new `Input` object. The aim is to make it harder to have errors like
https://github.com/sandstorm-io/capnproto/pull/270

This reduces dependencies on stdc++. For rationale, see
  https://github.com/kamalmarhubi/capnproto/commit/62a562f6d2c7bd147958106a1cf4169533d43f83#commitcomment-14375773

Requested by jo_liss on Twitter:

https://twitter.com/jo_liss/status/664884038252027905

(Also implement defaultValue<>() helper which was defined but apparently never implemented.)

Such values are not allowed by the JSON spec. We match the behavior of
JSON.stringify in Firefox and Chrome.

fixes https://github.com/sandstorm-io/capnproto/issues/261

Per spec.