Expose proof generation and verifying api.

[montekki]

Exposes proof generation and verification from trie-db.
Bumps trie-db and memory-db dependencies.

[NikVolf]

There should be PR in polkadot ready so that always identical version of trie* dependencies are used

[cheme]

Not sure if it should be part of this PR, but there is also : https://github.com/paritytech/trie/blob/d60e70bdd8c1e62af00459f9301c5002e69aaa21/trie-db/src/lib.rs#L105 that propose a way to have smaller proofs.
The method from this link targets proofs where we do not have content information (extrinsic execution or cumulus witness, or key value proof when we do not know the values).
It is still a processing / memory cost, and I am wondering if we should target support of multiple proof format. The method in this PR are specific enough that it seems very unlikely that someone would like to send a full proof instead.
I just realize that verification could be way faster if we verify against a previously checked proof (no need to recalculate the hash if they were already checked).
Just by curiosity what is the use case here (what I did understand from Jim was that it was mainly to check for change of values)?

[montekki]

@cheme For instance it is intended to be used in XCMP

[bkchr]

Spaces everywhere!

[bkchr]

Benchmarks need fixing.

[montekki]

@bkchr yes, that is in process

[rphmeier]

nit: "included into" should be "included in" - here and in other docs.

[rphmeier]

LGTM other than docs nit!