The current way we consider ER as confirmed is wrong

[NingLin-P]

Currently, the ER of #N is considered as confirmed whenever the ER of #(N + BlockTreePruningDepth) is added to the block tree:

https://github.com/subspace/subspace/blob/17225c4b10efb10fb0d4277de657236f32d0798b/crates/pallet-domains/src/block_tree.rs#L252-L263

Suppose the challenge period (i.e. BlockTreePruningDepth) of ER is 10 domain blocks, so the ER of #1 is considered confirmed when ER of #11 is submitted into the block tree.

But in the current implementation, the ER of #1 can be submitted any time before it is considered confirmed, so a malicious operator can submit an invalid ER of #1 when the head receipt number is #10 and right before ER of #11 is submitted. In this way, the invalid ER will be considered as confirmed before there is a fraud proof submitted and pruning it, then we will have 2 ERs of #1 considered as confirmed and will result in a MultipleERsAfterChallengePeriod error.

The correct ER should be considered as confirmed is the one that has 10 ERs built on top of it, but it is non-trivial to identify such an ER currently. It is also unnecessary because #1731 will fix the issue automatically by rejecting the invalid ER of #1 being submitted, if an operator wants to submit another ER of #1 it must submit a fraud proof to prune the current ER of #1 first.

cc @vedhavyas