docs(validation): refresh stale issue states — #50/#52/#57/#76/#78/#79 CLOSED

Six tracking issues the risk register / Layer-C narrative still asserted as
OPEN are all closed; each is updated to its gh-verified close reason, with the
historical finding and every honesty boundary preserved (no fix over-claimed):

- #50 (first-durable-commit granularity) -> FIXED by #91 (ebb573c): fetch-plane
  quantum warmup slow-start + resume-seeded discovery bound first-commit to
  seconds on the default-chunk path; commit plane / sync-store / schema untouched.
  Does NOT claim a fresh dense process's first discovery scan is warmup-bounded
  (that stays the expected, disclaimed property of section 4.4/5.5). Chaos
  discovery-geometry coverage via #92/#96.
- #52 (PGlite not crash-durable) -> CLOSED, superseded by #56 native-PG Tier 1;
  the WAL-tear finding stays a real, correctly-scoped Tier-0 limitation (section 5.4)
  -- closed because Tier 1 supersedes it for the acceptance evidence, not because
  PGlite became durable.
- #57 (partial persisted-finality floor) -> already fixed via timestamp-mapping;
  no residual OPEN assertion (mechanism rows left intact).
- #76 (upstream block.size off-by-one) -> CLOSED-as-tolerated via #77 (generalized
  in #107); the upstream dataset defect PERSISTS and is benign (non-consensus
  size field) -- not upstream-fixed.
- #78 (differ superlinear cliff) -> FIXED by #82 (dcc782a): cell diff defaults to
  the #72 paged/byte-aware differ.
- #79 (results-doc tag collision) -> FIXED by #80 (b1affee): unique seeded-random
  window tags across specs.

Also fixes a link nit (#92 is an issue, not a PR). Docs-only; still-open issues
(#23/#27/#53) left OPEN.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Author: dzhelezov