In-turn signers should override out-of-turn blocks

[soc1c]

We should add that in-turn signers should continue to produce their block even if they receive an out-of-turn block for that height before they publish their in-turn one. Otherwise clock drift and the lack of a minimum delay for out of turn blocks can cause an online in-turn validator from publishing their block and cause problems for the network. There's a more detailed explanation of how this happens at https://docs.google.com/document/d/1tmsr66sAPJmIZfSy5zck1uxnLzaXYxvzTNL5CF1pJQ0/edit#heading=h.ic4qz2ybx51

ref ethereum#1570 (comment)

[ChainSafeSystems]

Proposal: Add two configurable options for the Clique algorithm:

OUT_OF_TURN_DELAY_MULTIPLIER
Currently explicitly 500ms in the spec

MIN_OUT_OF_TURN_DELAY
New concept so Rinkeby and Görli are effectively using 0ms.
Absolute minimum amount of time to wait before publishing an out of turn block

Delay calculation is then:
MIN_OUT_OF_TURN_DELAY + rand(SIGNER_COUNT * OUT_OF_TURN_DELAY_MULTIPLER)

[soc1c]

9f5cfc8 proposes

• Each singer is allowed to sign maximum one out of SIGNER_LIMIT consecutive blocks. The order is not fixed, but in-turn signing weighs more (DIFF_INTURN) than out of turn one (DIFF_NOTURN). In case an out-of-turn block is received, an in-turn signer should continue to publish their block to ensure the chain always prefers in-turn blocks in any case. This prevents in-turn validators to be prevented from publishing their block and potential network problems.