The top table rule counts its chances over the wrong thing

Found while building TT-46, and separate from the coverage problem TT-47 and TT-48 deal with.

The top table rule counts its opportunities over the top-table seats that are occupied, rather than over the guest list. That contradicts the principle KB-8 states, and it means an empty top table has no opportunities at all and drops out of the mean rather than scoring nought.

It was invisible while the score covered soft rules only. TT-46 puts it in at weight three, so it is live now.

Acceptance criteria

• The rule's opportunities no longer depend on how many top-table seats are occupied
• An empty top table scores nought on this rule rather than being excluded from the mean
• KB-8 says what the denominator is, in the same breath as it says opportunities come from the guest list, so the next rule author has an example that does not contradict the principle
• The existing top table tests are re-derived against the new denominator

Open, for Rachel

KB-8 names no denominator for this rule and there are two candidates.

The top table's capacity. Each seat either holds the role the protocol gives it or it does not, which is the thing the rule actually checks. An empty top table then scores nought rather than being excluded, which is the behaviour wanted here.

The protocol-role holders on the guest list. Closer to the letter of "counted over the guest list", but a guest list holds eight protocol roles and a top table of six seats only six of them, so two are correct in being elsewhere and the denominator has to know that.

Recommendation is capacity, because it is what the rule can get right and it fixes the exclusion without a special case. KB-4 has the roles that do not fit sitting together at the nearest round table, which is a different check and arguably its own rule rather than part of this one.