note that std::numeric_limits<uint64_t>::max() and
std::numeric_limits<int64_t>::max() aren't exactly representable in a
double, so we need to be strictly less to be definitely lossless

UBSAN gave during Value.IsLosslessDouble test:
include/rapidjson/document.h:955:42: runtime error: value 1.84467e+19 is outside the range of representable values of type 'unsigned long'

maybe these tests should just be deleted?

UBSAN gave:
runtime error: signed integer overflow: -9223372036854775808 - 1 cannot be represented in type 'long'
runtime error: signed integer overflow: -9223372036854775808 - 2 cannot be represented in type 'long'

specifically, "expression with side effects has no effect in an unevaluated context"

Also remove SetArray|Object(…)

#330

Fix #341

As discovered by @felipegb94, there are missing overloads to the
`GenericValue::AddMember<T>` template function, taking an explicit
`GenericValue&` as a name and accepting arbitrary primitive values.

This patch adds the missing overloads. The `StringRefType` overload
is needed to disambiguate the addition of a string literal as
value.

Some tests are added to `TEST(Value, Object)` in `valuetest.cpp`.