Accept T2 by B1 mention check

[ggwpez]

Changes to a runtime API need a B1 label, but B1 rejects T2. Therefore:

• Accept the T2 label when B1 is specified

MR paritytech/polkadot#6536 and paritytech/cumulus#2073 suffer from this.
TODO:

• Add tests

[chevdor]

After some checking with Oliver, I think we can summarize this way:

The issue is that T0 and T1 are orthogonal: We can (according to our rules) have only one of them and we must have at least one of them.

Adding T2 is not an issue. But removing T1 while not having a T0 instead is not legal with the current rules.

@the-right-joyce it may be easier to move T0 and T1 away from the Ts and keep them with a letter that is exclusive.
That would also free 2 spots in the non exclusive Ts.

For now, it sounds like @ggwpez can live with adding back T1 on the mentioned PRs.

[chevdor]

There are 2 rules at play here and I missed that in my inital review:
https://github.com/paritytech/labels/blob/main/ruled_labels/specs_cumulus.yaml#L219-L231

Those say:

• if we have B1, we must have 1 and only 1 of the set [T0, T1]
• if we have B1, we must have 1..n of the set [T*]

We need to make both those rules happy.

[ggwpez]

Going to merge this now since I think its a valid use-case. However, in the MR that i mentioned it probably makes more sens to use both labels, since it does slightly alter the runtime blob.