tntcxx: update response decoding documentation

[drewdzzz]

Decoding and reading the data section is almost completely irrelevant since the decoder was reworked, we need to update it.

Class Data has decode method now. One can call it and pass a reference to container in which the tuples are to be placed. For example, your tuples has format [int, int]. That's how decoding of such tuples looks like:

using MyTuple = std::tuple<int, int>;
std::vector<MyTuple> tuples;
data.decode(tuples);

Generally, here are supported objects:

• std::nullptr_t will be always decoded as MP_NIL.
• fundamental types (int, float, bool, ...)
• char[N], std::string for MP_STR.
• Any putable, contiguous (if underlying type is not char) or tuple-like object is considered as MP_ARR.
• Any putable, contiguous (if underlying type is not char) or tuple-like object, containing pairs (including std::map), is considered as MP_MAP.
• Any optional-like object can be decoded as MP_NIL or as its underlying value.
• Any variant-like object can be decoded as any of its alternatives.

Also, one can populate its own class with encoding rule using static constexpr mpp_dec or mpp member, containing pointers to the class member. Used members must be public. And, one can use mpp::as_* tag to specify what type to encode into. Semantics is the same as the one in #3967 (these features are described in the end of the issue).