pallet-migrations: try-runtime support, and manage author mapping blake2 migration

[notlesh]

What does it do?

This adds try-runtime support in pallet-migrations so that the feature can be used as it would in a normal migration.

What important points reviewers should know?

This PR currently uses notlesh-migration-sketch branch as the target for merging. I did this so we have the option of merging the other without this one.

Is there something left for follow-up PRs?

The motivation for this is, in part, to allow #679 to use both try-runtime and pallet-migrations.

TODO:

• Write test to cover making sure post_upgrade() only runs migrations that weren't skipped (see comments)
• Fix cargo test, currently requires cargo test --features=try-runtime (see Activating features for tests/benchmarks rust-lang/cargo#2911)

[notlesh]

This is building and appears to be working now (with Joshy's author_mapping migration).

I was able to test with try-runtime against moonriver, but I needed to make a small mod to substrate:

diff --git a/utils/frame/remote-externalities/src/lib.rs b/utils/frame/remote-externalities/src/lib.rs
index 53c44780a6..8d1a686608 100644
--- a/utils/frame/remote-externalities/src/lib.rs
+++ b/utils/frame/remote-externalities/src/lib.rs
@@ -41,7 +41,7 @@ type KeyPair = (StorageKey, StorageData);

 const LOG_TARGET: &str = "remote-ext";
 const DEFAULT_TARGET: &str = "wss://rpc.polkadot.io";
-const BATCH_SIZE: usize = 1000;
+const BATCH_SIZE: usize = 100;

 jsonrpsee_proc_macros::rpc_client_api! {
    RpcApi<B: BlockT> {

Without this, I found that the requests would consistently time out at the same batch. This was when using a locally-synced archive node (requests going over loopback).

[girazoki]

We are only using this in the crate context, we can make it pub crate only

[JoshOrndorff]

That's a valid point, but our other pallets also have them just pub (eg https://github.com/PureStake/moonbeam/blob/master/pallets/asset-manager/src/lib.rs#L119). And I worry that making stuff too restricted will make it hard to write precompiles in the short term.

Here's a followup issue to treat this throughout the repo. https://purestake.atlassian.net/browse/MOON-933