warp: add "requirePrimaryNetworkSigners" precompile config

[darioush]

Why this should be merged

#1230

How this works

Adds a precompile config option where "requirePrimaryNetworkSigners" can be added at warp activation (or de-activated and then re-activated with this flag by existing networks).

Note we can't use a "per node" config, since when the optimization is applied the validator bitset is modified.
Basically the existing rules do not allow for the primary network's validators to sign for themselves but expects only the subnet validator set. Since the entire validator set is replaced, this includes expectations for the bitset, etc.

How this was tested

CI:

1. Expands UT for TestWarpMessageFromPrimaryNetwork (in precompile/contracts/warp/predicate_test.go) to include both cases for requirePrimaryNetworkSigners,

2. Modifies (expands) TestReceiveWarpMessage:

• first asserting warp messages can be received as expected with the default config (no RequirePrimaryNetworkSigners) for messages from subnets, non-P chain primary and P-Chain primary, this ensures no regression for existing networks,
• disabling then re-enabling the warp configuration with RequirePrimaryNetworkSigners (note disabling is required as it is not allowed allowed to update the precompile config)
• next asserting its behavior again with the RequirePrimaryNetworkSigners (non P-Chain primary network chains should use primary network).

I also added a check the result of the predicate in the block since the previous test asserts behavior about whether the block is valid or not in a different height context, but the block would be considered valid (but failed warp message), then the assertion that does a trace comes later.

It seemed preferable to verify the predicate results are as expected instead of re-ordering the existing code.

How is this documented

Not yet

[ceyonur]

overall LGTM, just a nit to convert the function to an interface

[ARR4N]

How this was tested

CI

I think the intention of this block is an explanation of what tests are being performed. I see that you made changes to the unit tests but I don't have the context to understand why they are sufficient to prove that this works.

[ARR4N]

LGTM. Thanks for the changes.

Caveat: I don't know enough about the specific test cases in the table nor the implication of mySubnetID in validators.NewState() to comment on them, but I think it's all now clear enough that I'm deferring to Cey's existing Approval.

[michaelkaplan13]

Makes sense to me 🙏