service/header/core_listener: Fix bug where old block headers are spammed to headersub by bridge nodes

[renaynay]

Resolves #499.

[adlerjohn]

If there are still TODOs, can we convert this to draft?

[renaynay]

How do you turn it back to draft mode @adlerjohn ?

[adlerjohn]

[renaynay]

TestSyncStartStopLightWithBridge is flakey on main branch as well.

[Wondertan]

Appendix to my review. The assumption about fixing the issue has to be verified somehow. Manual verification is also fine and shouldn't be hard to setup

[renaynay]

Core_listener should still listen for new blocks while bridge node syncing and erasure code them but should not broadcast them yet.

[liamsi]

This is a straightforward change and PR. Thanks!

[liamsi]

Only thing that could be improved is related to:

Core_listener should still listen for new blocks while bridge node syncing and erasure code them but should not broadcast them yet.

Ideally this was captured by a test somehow (unit/swamp).

[Wondertan]

We found an issue with the approach there and after discussing plenty of solutions decided to change the solution here. Specifically, we decided to contribute to the PubSub for a specific feature that would allow us to solve the problem: libp2p/go-libp2p-pubsub#481
We won't wait for it to be accepted as we maintain our own PubSub fork due to #306, so I am cherry-picking the solution now and making an alternative PR. If something is change on main pubsub branch, I will do changes here acrrodingly.

[Wondertan]

Closed in favour of #549