feat: notify dummy miner for new work

[quake]

What problem does this PR solve?

We often use the dummy miner mode for local development, but currently the dummy miner only submits new blocks at configured intervals, when other nodes broadcast the new block, the dummy miner will not switch to the latest tip but continue to submit the old block, so it will become an uncle block.

This behavior is fine in single-node mode, but when we need to simulate the hash rate distribution of multiple nodes for testing, there is no way to simulate this behavior with the dummy miner.

What is changed and how it works?

Tweak the dummy miner inner loop, if there is new tip (pow_hash changed), start working on the new one.

Check List

Tests

• Manual test (add detailed scripts or steps below)

# init two nodes
./ckb init -c dev --genesis-message dummy-miner-test  --ba-arg miner1-arg
./ckb init -c dev --genesis-message dummy-miner-test  --ba-arg miner2-arg

# add nodes to the whitelist section and config two miners' ckb-miner.toml
# for example, miner1 is 5 sec , miner2 is 9 sec
value = 5000
value = 9000

# start ckb nodes and miners and verify the block cellbase distribution by rpc `get_block_by_number` or check the log manually, it should be distributed as 5 / 14 =~ 35.7% and 9 / 14 =~ 64.3%

Release note

Title Only: Include only the PR title in the release note.

[eval-exec]

How about this:

use crossbeam_channel::select;

...


    fn solve(&self, pow_hash: Byte32, work: Work, nonce: u128) {
        select! {
            recv(self.worker_rx) -> msg => {
                if let Ok(WorkerMessage::NewWork {
                        pow_hash: new_pow_hash,
                        work: new_work,
                        ..
                }) = msg {
                    if new_pow_hash != pow_hash {
                        pow_hash = new_pow_hash;
                        work = new_work;
                    }
                }
            },
            default(self.delay.duration()) => {
                if let Err(err) = self.nonce_tx.send((pow_hash, work, nonce)) {
                    error!("nonce_tx send error {:?}", err);
                }
                return;

            }
        }
    }

[quake]

The current develop branch use a simple thread sleep, and I wanted to keep the code as unchanged as possible, rather than introduce a tokio runtime.

[eval-exec]

I wanted to keep the code as unchanged as possible

OK

rather than introduce a tokio runtime.

But self.nonce_tx's type is crossbeam_channel::Sender. We won't need tokio runtime here.

[quake]

I see, I thought it's tokio's select macro, let me change it and test later, thanks.

[quake]

after test, I found the code generated by select macro is different, the pow_hash and work vars are not read again, I will keep current code.