docs: record the corpus shape measurement in Field Note 05

[carlok]

The note argued all day about whether the boundary works and never asked whether the corpus accumulates. That question now has a number on the catalogue, and the note should say what it is and why it is measured that way.

New section, placed before the close:

• 88 of 99 accepted statements mention a definition existing only here, which sounds strong and is mostly self-reference — a submission proving things about the definition it just introduced. It rules out verbatim restatement and says nothing about accumulation.
• The graph is what answers it: 24 modules, 4 internal import edges, 0.17 per accepted submission; 26 corpus constants referenced in statements, 4 of them by more than one submission. Four small clusters, sixteen modules alone.
• Two measures because they are not equally honest, with the gaming asymmetry stated where a contributor reads the ratio.
• Why it is generated rather than written: a metric published only when it flatters is not evidence.

The close now opens on three questions rather than two, noting the first is at least counted. If the ratio climbs the premise has something behind it; if it stays near one edge per six, what exists is a growing collection of correct independent libraries — a fine thing, and a different claim from the one on the front page.