RuntimeInfo is not generic

* master:
  Revert "Install llvm-tools-preview component (#643)" (#644)
  Revert "fix wasm executor compile error (#631)" (#642)
  Fetch parity-common dependencies from crates (#617)
  limit retry count in OnDemand (#513)
  Minor refactor for staking module (#659)

This reverts commit ad89579f.

This reverts commit 7c18bb60.

* Fetch parity-common dependencies from crates

* Update triehash

* Fix code to fit latest triehash
Fetch kvdb-rocksdb from crates

* type params

* lockfile

* lockfiles

* wasm binaries and lockfile

* Somerhing wrong.

* My attempt to fix

* cfg_attr for serde

* Fix tests

* master:
  Upgrade to libp2p master (#660)
  Include function comments into modules `json_metadata` (#657)
  Replace old headers with CHT in light clients (#512)
  Fix build
  Introduce treasury and document (#646)
  Off-the-table staking preference (#656)
  Implement function `json_metadata` in `decl_module!` (#654)
  Fix warnings in networking (#652)
  Add a reputation system (#645)
  Check for pruned block state (#648)

* Include function comments into modules `json_metadata`

All doc comments (with `///`) are included in the description field of
the function.

Progress on #535

* Replace spaces with tabs

* storage proofs

* CHT

* Introduce treasury and document

* Revert bad changes

* More reversions

* Add example crate

- Remove HasPublicAux
- Rename Concrete -> Runtime

* Actually commit stuff

* Changes

* Propagate block number in finalise.

* Fix and build example

* Fixes.

* Fix compilation for treasury.

* Fix the treasury test

* Tests

* Fix.

* Fix tests

* Fix a few grumbles

* Fixes

* Fix grumbles

The new function `json_metadata` returns the module declaration
as a json string.

Progress on: #535

* Add a reputation system

* Fix tests

* Don't try to dial peers to which we are already connected

* Use the master branch of libp2p

Constrain the wasm-version using an ExternTrieCrypto trait that is implemented only for KeccakHasher

Merge remote-tracking branch 'origin/master' into dp/make-CallExecutors-generic-over-hasher-and-rlpcodec

* origin/master: (26 commits)
  Contract runtime polishing (#601)
  WIP on chain heap (#639)
  Events to track extrinsic success (#640)
  Install llvm-tools-preview component (#643)
  fix wasm executor compile error (#631)
  random fixes (#638)
  Empty becomes (), reflecting convention (#637)
  Allow to build_upon skipped entries, but don't walk back (#635)
  Separate out staking module into balances and payment (#629)
  Update .gitlab-ci.yml (#633)
  Do not attempt to rustup if in CI. This is taken care of by the base (#621)
  Avoid need for ident strings in storage (#624)
  rename to panic_handler as panic_implementation is deprecated in nightly (#626)
  5 random fixes (#2) (#623)
  fix one typo in README (#627)
  Misspelled words (#625)
  Contracts: Per block gas limit (#506)
  Make sure to ban invalid transactions. (#615) (#620)
  Forward-port BFT fixes from v0.2 and restructure agreement cancelling (#619)
  Allow specifying listening multiaddresses (#577)
  ...

* Rename ext_transfer → ext_call

There are several tests where gas was changed. This is due to shrinking
the name and consequently a size of contracts.

* Fix macros

Specifically, make fixes that allows to use a function without arguments (apart from ctx).
For example:

```rust
ext_return_42(ctx) -> u32 => {
    Ok(42)
}
```

Also, add impl ConvertibleToWasm for u64.

* Implement passing input data buffer from top-level

This allows the transaction initiator to pass some input data that can be accessed from the executed contract.

However, this only works for the top-level contract. The contract still cannot pass input data to the callee contract. The reason for that is that API doesn't support this presently.
This will be addressed in the following commits.

* Input data passing thru `ext_call`

`ext_call` now takes a buffer (or more specifically, pointer  to the start of and size of a buffer).

This allows contracts to pass input data while calling into other contracts.

* Input data passing thru `ext_create`

Same as previous commit but for `ext_create`.

* Trap on decoding error instead of panicking

* Don't cascade failure in ext_call

Instead, if callee doesn't complete successfuly, return a non-zero status code.

This will prevent cascading traps up to the top-level. Due to this some tests were altered so that they now expect successful transfer instead of a failure.

* Don't cascade failure in ext_create

Same as previous commit but for ext_create.

* Specify how much gas to allot for a call

* Specify how much gas to allot for a create

* Update tests.rs

* move heap size on chain

* fix the interface change

* decode heap size

* fix code comments

* fix comment

* update Cargo.lock

* rename to heappages

* add one heap pages variable in runtime

* Initial draft

* Fixes

* Extrinsic success events