BIP-431: clarify Rule 3 to enforce cluster size 2 and fix inclusive counting

[MozirDmitriy]

This change resolves an internal inconsistency in BIP-431 between Rule 3 and the “Related Work” section (and linked Core PRs) that assume clusters of at most size 2 for TRUC. Previously, Rule 3 allowed up to one unconfirmed ancestor and one unconfirmed descendant simultaneously, which permits a 3-transaction cluster (parent → tx → child). However, “Related Work” states “1-parent-1-child (aka cluster size 2)” and references features that rely on maximum cluster size 2 (feerate diagrams: PR 29242; package RBF: PR 28984; sibling eviction: PR 29306).

what is the problem:
Rule 3 allowed “≤1 ancestor” and “≤1 descendant” at the same time, enabling a chain of three unconfirmed transactions.
“Related Work” asserts “1-parent-1-child (aka cluster size 2)” and the referenced PRs constrain features to clusters of size ≤2, which is incompatible with a size-3 cluster.

[murchandamus]

There is no inconsistency here. Transactions being limited to a single ancestor and a single descendant obviously limits TRUC transaction clusters to two transactions.
This is already explicitly explained in the BIP, and the proposed changes here do not improve the BIP.

I am noticing you as a repeat author of trivial PRs. Please do not open pull requests unless they constitute a significant improvement.