Encapsulate access to the client for secret store

[grbIzl]

Hopefully it's the last step, that will allow to separate SecretStore to its own repository.
Currently Secret Store uses Client in order to retrieve information from the chain, mostly for its service contracts (like check of permissions). This functionality should be hidden behind abstract trait, that should be implemented inside ethcore.
Details about the solution:

Before this change:

• SecretStore server is started (from parity main) using method with parameters-pointers to Client, Miner etc:
  pub fn start(client: Arc<Client>, sync: Arc<dyn SyncProvider>, miner: Arc<Miner>, self_key_pair: Arc<dyn NodeKeyPair>, mut config: ServiceConfiguration, db: Arc<dyn KeyValueDB>, executor: Executor) -> Result<Box<dyn KeyServer>, Error>
• SecretStore module has build dependencies to ethcore libraries
• Parity module has thin layer inside to start secret store (mostly with required parameters)

After this change:

• SecretStore server is started (from parity main) using method with parameter-pointer to SecretStoreChain structure, declared in SecretStore module:
  pub fn start(trusted_client: Arc<dyn SecretStoreChain>, self_key_pair: Arc<dyn SigningKeyPair>, mut config: ServiceConfiguration, db: Arc<dyn KeyValueDB>, executor: Executor) -> Result<Box<dyn KeyServer>, Error>
• This newly added structure doesn't have dependencies to ethcore in its API. So, as a result, SecretStore module doesn't have any dependencies to ethcore or any other internal Parity libraries. It has only dependencies to public crates from crates.io
• Parity module has thicker layer inside to start secret store, that contains implementation of SecretStoreChain (this implementation has dependencies to ethcore, but as it's the internal implementation inside parity, it's ok)

Tasks:

• Move client related calls from all SecretStore modules to TrustedClient
• Rework\duplicate used ethcore\common_types
• Rename TrustedClient to (?) ClientWrapper
• Move TrustedClient implementation into ethcore and leave only trait declaration for SecretStore
• Check tests
• Wrap Registar and CallContract interfaces
• Solve parity runtime dependency - will be resolved separately after Parity runtime moved to parity common for publication in crates.io paritytech/parity-common#271

[dvdplm]

* Solve parity runtime dependency

This one could be pretty involved I think. Do you have any thoughts on it (yet)?

[svyatonik]

Thanks! LGTM, though most of introduced traits will be replaced by higher-level traits later. Like some new concepts are still ethereum-specific and won't allow us to use SS in other projects. E.g. contracts are still call Blockchain::retrieve_last_logs() and manually parse ethereum logs. But imo this should be merged as is, because it allows us to extract SS to separate repo (which is, again a necessary step to use it in other projects).

[ordian]

Merge conflicts need to be resolved and I thought paritytech/parity-common#271 needs to be merged first?

[grbIzl]

I thought paritytech/parity-common#271 needs to be merged first?

We discussed it @dvdplm and decided to try to separate secret store code into the new repository locally, in order to make sure, that we don't need anything except runtime (that is in PR 271). But in order to proceed to this, I need this preparation PR to be merged.

[dvdplm]

Left a few comments, but I'm too unfamiliar with the secretstore code to do a proper review. I am also a bit unsure of what the path looks like from this to a stand-alone secretstore; maybe you can extend the PR description to help reviewers?

[dvdplm]

To @svyatonik's point about avoiding "ethereum-isms": should we provide a default impl here?