Improve nonce handling

[echevrier]

Currently when we use the extrinsic factory to create extrinsics we use a cached nonce value. As we do not always know when an extrinsic has not been executed successfully, all subsequent extrinsics will have a wrong nonce. (See issue).

Furthermore, what happens when the extrinsics are sent in parallel? Can we ensure that the nonce is unique?

Answer @haerdib : Yes, because each enclave has its own, unique signer and the nonce cache is behind a lock. So if the ExtrinsicFactory is accessed for each extrinsic creation, the nonce is unique.

How do we handle extrinsics that have already been sent with the wrong nonce?

What other issues should be considered?

[echevrier]

Currently, the cached nonce is not systematically updated. See #1007 (comment)

[murerfel]

One step towards solving this, as discussed in #1033, is to ensure that we use the ExtrinsicsFactory whenever we want to compose extrinsics. We still have a couple of places (e.g. attestation), where we duplicate the extrinsics factory code (and sometime forget to increment nonce, which is handled consistently by the factory)

[haerdib]

I will try to solve this while doing #1049

[ziming-zung]

After merging #1054 , i tested the issue and found that the issue has not been resolved. Can you help to double check?

[haerdib]

Hm.. you're right, I guess. While hopefully all extrinsic offline creations now use the ExtrinsicFactory, in case of a nonce mismatch ( a call lost somewhere or invalid) the nonce will not be correct anymore in the local storage.

@clangenb please correct me if I'm wrong.

However, there's already a similar issue: #231 which I just updated accordingly.

Edit:
Sorry my bad, #231 is actually unrelated, as it affects the cli, not the nonce storage of the enclave. I guess this issue might need to be reopened, or is there a similar issue already?

[Kailai-Wang]

@haerdib maybe this one: #970