Reduce load on validator subscription channels

[pawanjay176]

Issue Addressed

N/A

Proposed Changes

Since we now only subscribe to a fixed number of long lived subnets per node, the validator_index field in ValidatorSubscription type is entirely inconsequential to the beacon node.

This PR does 2 things:

1. Removes the validator_index field from the ValidatorSubscription type. This allows us to filter duplicates which can be quite a lot in high validator count setups.
2. Instead of sending each subscription from the http api to the network service in a separate message, we send it in one single message.
   I'm not completely sure at what size of the vector we'll start seeing slowness in the channel, but by filtering duplicates, we shouldn't be sending too many.
   Max length of the vector would be max_committees_per_slot * slots_per_epoch * 2(aggregator bit) = 4096 .
   Not super attached to this change, but if we send one by one, then we should probably change the ValidatorSubscriptionMessage to take just a single message instead of a vector for consistency.

[AgeManning]

Yeah nice. Reducing messages ftw

[dapplion]

LGTM

[pawanjay176]

@michaelsproul Can we roll this in with #5305 ?

[michaelsproul]

Looks good with the BTreeSet tweak!

Happy to merge and continue testing on infra over the weekend

[pawanjay176]

@Mergifyio queue

[mergify]

queue

✅ The pull request has been merged automatically

The pull request has been merged automatically at 84a902a