feat: delayed finalization

[evilrobot-01]

Adds --finalize-delay-sec cli argument, based on the great work by @shunsukew at inkdevhub/swanky-node#61.

Manual testing by starting node with/without option, along with example contract upload:

# build example contract
cargo contract build --manifest-path=../ink-examples/erc20/Cargo.toml

# start node
cargo run

# upload contract
cargo contract upload --suri //Alice --execute --manifest-path=../ink-examples/erc20/Cargo.toml
# check finalized head remains at genesis
sleep 1
test $(curl -sH "Content-Type: application/json" -d '{"id":"1", "jsonrpc":"2.0", "method": "chainHead_unstable_genesisHash", "params":[]}' http://localhost:9944 | jq .result) \
  = $(curl -sH "Content-Type: application/json" -d '{"id":"1", "jsonrpc":"2.0", "method": "chain_getFinalizedHead", "params":[]}' http://localhost:9944 | jq .result) && echo PASS || echo FAIL


# start node (with delayed finalization)
cargo run -- --finalize-delay-sec 1

# upload contract
cargo contract upload --suri //Alice --execute --manifest-path=../ink-examples/erc20/Cargo.toml
# check finalized head matches chain head
sleep 1
test $(curl -sH "Content-Type: application/json" -d '{"id":"1", "jsonrpc":"2.0", "method": "chain_getHead", "params":[]}' http://localhost:9944 | jq .result) \
  = $(curl -sH "Content-Type: application/json" -d '{"id":"1", "jsonrpc":"2.0", "method": "chain_getFinalizedHead", "params":[]}' http://localhost:9944 | jq .result) && echo PASS || echo FAIL

Closes #160

[athei]

Thanks a lot. We have to keep in mind that the default behaviour changes now from "never finalize" to "finalize right away".

I think it would be better to change the default value so something like 2 seconds instead of 0. It is a more sane default as it forces client devs to think about those two states when working with the dev node. Instead of assuming they are the same.

cc @wottpal

[evilrobot-01]

It is currently opt-in only by specifying the --finalize-delay-sec option, so the value is left up to the user. The default behaviour should therefore remain the same for any existing use cases unless explicitly specified, and manually specifying it should force the developer to think about the additional state. I can see how having it as a default might be useful, but I thought it better to start with it as optional.

Perhaps the notes in the command description/readme around using 'a value of 0 for instant finalization' should simply be replaced with a suggestion of 2 seconds? The command example in the readme suggests 5 seconds, so I can standardise all of the documentation to 2 seconds.

[wottpal]

thanks for the heads up, @athei. really appreciate that this is coming to substrate-contracts-node finally. 🚀 no strong opinion about making it the default or not. but if yes, something like 1s would be totally sufficient to make frontend devs aware of handling the additional state IMO.

[athei]

I did misread the code. Didn't notice it was an Option. So yes indeed the default behaviour does not change with this PR. This is okay as it is non invasive.

However, I don't think having blocks to be never finalized is of any use (current behaviour). I suggest removing the Option and just set the default to 1 as @wottpal suggested. It should still be okay as we only add finalization. e2e tests listen for block inclusion and shouldn't be slowed down by this.

[evilrobot-01]

Updated as suggested, where it now defaults to delayed finalization of 1 second but still allows developer to override.

Note that the first manual test in the original PR comment above no longer applies, as that was testing the default method remained unchanged.