trie: add ChildTrie trait to simplify future child trie types addition

[sorpaas]

This refactoring adds ChildTrie trait. The aim is to simplify child trie types addition.

With this PR, to add a new child trie type, one only needs to:

• Create a new file under core/trie/src/child_tries, adds a new unit struct, and implement ChildTrie.
• Modify the with_child_trie macro, add a new if condition defining the new prefix.

[gui1117]

tests fail

[cheme]

looks good, maybe as small doc around the macro to make it goal explicit could be good (resolving child trie implementation from storage_key).

[cheme]

I wonder if we may rename ChildTrie with something a bit more generic, I would say ChildState but that is not really related to the following methods either, so here to me KeyValueState is what makes the more sense.
Similarily we can remove the 'child' info here: there is nothing related to the child nature.
Still If we add method or associated const such as

  /// ChildTrie storage key defines its type. eg: the default child trie is using
  /// well_known_keys:: CHILD_STORAGE_KEY_PREFIX followed by ':' then
  // 'default', default being this associated const.
  const TYPE_PREFIX: &'static[u8] =  b"default:";

or

 get_type_prefix() -> &'static[u8];

[cheme]

for all those function I would consider removing type parameter L, and pushing it into the struct

struct DefaultTrie<'a, L, DB, Query>{
  root: &'a[u8],
  _ph: PhantomData<(L, DB, Query)>
}

instead of

struct DefaultChildTrie;

Actually this comes with some constraint (probably associated type for prover and and memorydb), so it may not be worth it at this point.

[cheme]

Note that it will be good to have a constant reference instead of b"child_storage:default:" and if we put a function/constant in ChildTrie trait to define its type prefix, it should also be use.

[gavofyork]

Useless docs.

[gavofyork]

Useless docs.

[gavofyork]

Useless docs

[gnunicorn]

is still being worked on?

[sorpaas]

Sorry. Let me close this for now.