fix: include genesis validators in gateway parent validators tracker

[sanderpick]

This PR is the result of trying to track down the cause of the following behavior:

• Create a new subnet with min validators = 2
• Create 3 validators and join
• The first 2 validators will be considered "genesis validators", and the third one will be added to the power table (membership in the subnet gateway contract) via a top down staking change
• The subnet receives and stores the validator change, then checkpointing applies those changes. The next power table excludes the first two validators here. We now have only one validator with power.
• Since the genesis validators are not included in the subnet gateway's parent validator tracker, the new membership set by the top down finalizer excludes the genesis validators, ie, listActiveValidators() will be zero when applying this first change here.
• Bottom up checkpoints fail because they are only signed by this one validator with power, but the subnet contract on the parent still thinks (correctly) that the total collateral is from three validators, ie, the quorum threshold is higher than what's included in the checkpoint. Checkpoint fails with WeightsSumLessThanThreshold
• You can see this reflected in the power as known by cometbft by looking at the nodes' status endpoint (http://localhost:<comet_rpc_port>/status). Power is zero for the first two nodes.

I'm not sure if this fix is ideal, but it works. It just adds the genesis validators to the parent validator tracker before setting the initial membership in the gateway constructor.

[sanderpick]

Hunch: This issue is also the reason that in federated mode, you have to always include all validators, instead of being able to add them one-by-one.

[sanderpick]

Ah, makes sense. Somehow I missed https://github.com/consensus-shipyard/ipc/issues/1150. I'm happy to take a stab at passing the mode in the constructor here, but sounds like from slack convo you're picking this up. If so, do you want access to this branch or will you just recreate it? Nbd either way.

[cryptoAtwill]

@sanderpick yeah, it's quite tedious to fix the constructor, quite some code needs to be updated in fendermint. @raulk mentioned yesterday that he will check this out to see if there are easier fixes. But I think for collateral based subnets, this is the fix.