Improve reflection system and static_variant

[nathanielhourt]

Previously, fc reflection generated visitor infrastructure capable of
visiting each field of a struct; however, this is only useful for run-
time operations on structures. No compile-time information was preserved
about the fields of a struct, making compile-time introspection of types
impossible.

Now FC reflection generates rich compile-time information about the
members of structs, including types, names, and structure, allowing
code to be written to explore types in detail at compile-time and fetch
arbitrary fields at runtime without needing to iterate over the unwanted
fields.

To make this a reality, a new typelist type was added to store this
compile-time information. This type is also useful within the context of
static_variant, as SV previously used implementation detail types in a
private namespace to provide equivalent functionality. Now static_variant
uses the reusable functionality of typelist to operate, dramatically
reducing the amount of code dedicated to static_variant and also
making static_variant types more flexible to work with since their
infrastructure is now based on the typelist public interface.

These changes are necessary for bitshares/bitshares-core#1860

[pmconrad]

Ah, you make this look so simple! Excellent work! :-)
@abitmore?

[abitmore]

@pmconrad I haven't fixed the TRAVIS_BUILD issue yet, it does not only affect Travis though.

[nathanielhourt]

Ah, you make this look so simple! Excellent work! :-)

Thank you very much, Peter =) That means a great deal to me.

[nathanielhourt]

Shiny new static_variant::operator== in a75fcb5

I'm opting to leave out a default operator< as it's not immediately obvious how to resolve comparison between unmatched types. Not to mention, our flat_set usages depend on the fact that (static_variant<int>(1) < static_variant<int>(2)) == false. type_lt gets us that, but it's dangerous to add a default operator< in light of this fact.

[pmconrad]

Please rebase -i and fixup the travis workaround removal.

[nathanielhourt]

Done

[pmconrad]

The fixup went into the wrong commit, but nevermind.