Record the microVM, floor and per-VM-memory state at wind-down - #11934

Written on operator instruction to wind down and record remaining items, so
that v1 performance and v2 migration can be prioritised. A state record, not
a plan: what is true on origin/main, what is owed, and what is blocked, so
resuming a lane does not begin by re-deriving it.

One new file under docs/plans/. No code, no model changes, no generated
artifacts.

Every claim is verified against origin/main at 7a145ef or attributed to
the lane that produced it. The floor's per-phase fail-open reading is marked
CONTESTED, carrying both readings and the discriminating test that would
settle it, rather than asserted in either direction (DESIGN 4d: an inference
is typed as a bet, and the reader who consumes it as a fact is the defect).

Corrects three premises of my own closeout instructions, each caught by the
lane it misdescribed: both dynamic-memory branches COMPILE and run green
(the uncompiled attempts are the two closed PRs), #11751 is in the merge
queue rather than parked, and both parked PRs were non-draft.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Both findings verified and both hold.

TRANSCRIBED MEASUREMENTS (DESIGN 6). Every copied tally is replaced by the
producer that re-derives it: the floor census now names the required floor
invocation that prints its own summary line rather than quoting counts that
move with every landed witness; the floor-fit figures name
gunbc.runner_slot_allocation and gunbc.floor_memory_demand and send the
reader to the stall row where the numbers sit beside the reading that
produced them.

Where a figure is a lane's reported reading rather than a re-derivable
instrument -- the srv1 wet run, the two branch claim_batch results -- it is
now ATTRIBUTED as a lane receipt of a past run, with the re-derivation named,
instead of being stated as a standing fact. That distinction is the point:
those are receipts, not facts a reader may refresh, and they are written so
staleness is visible rather than plausible.

UNLINKED PLAN (DESIGN 3c). The finding is correct and the page is an orphan.
Rather than hand-add a gunbc.design_document row -- DESIGN.md is a projection,
and editing that authority is not a wind-down act -- the page now names its
consumer (the two re-entry briefs that cite it by path) and DECLARES the
frontier with its trigger, which 3c admits. It also admits by name that it
joins the population already rostered as doc_graph_orphan_population_stall
rather than adding to it silently. Trigger: whichever funded lane resumes a
program named here links this page from the section governing its work, or
supersedes and deletes it.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>