docs(readme): the lineage line claims both halves, because both are load-bearing

"Almost none of this is new" opened by conceding the weakest reading of the
ledger and then hedging it back. The accurate claim is stronger and is two
claims: fifty years of replicated software-engineering results, AND research
published last month.

The line now reads "Fifty years of software-engineering results, and research
from last month -- McCabe (1976) through to seven papers published in the last
two months." Both ends are real and both are checkable in docs/LINEAGE.md.

The recency numbers are derived, not asserted. Dates came from the arXiv API,
not from the ID prefix, which does not track publication date: 2607.09691 was
published 2026-06-19 despite its 2607 stem. Counting IDs would have put eight
papers in the window; the real figure is seven.

  17 papers from 2026 | 7 in the last two months | 3 in the last thirty days
  (as of 2026-09-08; every row carries its arXiv id)

The body now says why both halves are needed rather than apologising for the
old one: the classics are what make the quality lens trustworthy, the 2026 work
is what makes it current, and neither alone would be enough. It closes on the
newest row being a result that FAILED when tested here, which is the reason the
ledger is worth keeping at all.

NOTE, unguarded: "seven" and "three" are dated counts no gate re-derives.
readmedriftcheck holds 42/67/237 to LINEAGE's tables; these two are hand-checked
and will drift as papers age out of the window.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>