rbtree::emplace_hint() compilation fix for variadic arg use

[miherius]

This PR provides a fix for the rbtree::emplace_hint() method, when it called using multiply variadic template arguments. The problem is in DoInsertValueHint() method, which can't accept the forwarded parameter pack, expecting an already constructed pair only. A small example:

eastl::map<int, char> m;
m.emplace(1, 'a'); // ok, arguments are forwarded to the `pair` constructor
m.emplace_hint(m.end(), 2, 'b'); // compilation fail, arguments can't be forwarded
m.emplace_hint(m.end(), eastl::make_pair(2, 'b')); // ok, passing a constructed `pair`

This PR makes the interface of the DoInsertValueHint() method (internal helper for "hint" versions of insert methods) similar to the DoInsertValue() method (another helper for non-"hint" insert methods), so it can be used with variadic template arguments now.

Additional fixes:

• rbtree::try_emplace() move rvalue-ref key instead of copy;
• rbtree::try_emplace() delegates value_type construction to the internal logic;
• "hint"-version of non-unique hashtable::insert() can be complied now (was needed for new tests).

Map and multimap tests have been extended to cover more use-cases of these methods (changes in TestMapCpp11(), TestMultimapCpp11() functions).

There are no changes in the public interface of the mentioned classes.

[rparolin]

Thanks for the PR! Looks good to me.