docs: remove superseded repository archive copies

[forkwright]

Finding

docs/archive/README.md and docs/archive/STANDARDS.md are superseded
copies whose durable content is already owned by living documentation and Git
history. Their manifest entries preserve a stale archive-retention rule and
create a second apparent authority.

A clean local commit exists at 5912f25 on
docs/411-remove-superseded-doc-archive, based on current Akroasis main.

Done when

• Kanon issue owning the upstream Akroasis planning archive smelt is merged
  first, so the corpus has one consistent archive posture.
• The two repository archive files and only their manifest entries/comment
  are removed; all unique requirements remain owned elsewhere.
• A fresh exact-tree review, public Akroasis GitHub CI, merge-main checks,
  and normal branch/worktree cleanup complete.
• No research corpus is deleted in this change; its Phase 03 ownership is
  reviewed and landed separately.

Boundary

Git history remains the archive. This issue authorizes no hardware activity and
does not include the still-separate research/ retirement.

[forkwright]

Status check against current state, so this is not re-derived or swept by
mistake.

The work exists and is unpushed. 5912f25 sits in the worktree
projects/worktrees/akroasis-doc-archive on branch
docs/411-remove-superseded-doc-archive, based on e6cff1e. The tree is clean
and the branch has never been pushed to origin. Its diff is 85 deletions
across exactly the three paths this issue names — docs/MANIFEST.toml,
docs/archive/README.md, docs/archive/STANDARDS.md — and both archive files
are still present on main, so the work is genuinely outstanding rather than
already landed under another commit.

It is still gated. The first Done when clause requires the kanon issue
owning the upstream Akroasis planning-archive smelt to merge first, so the
corpus keeps one consistent archive posture. That issue is kanon#3555, and it is
open. Nothing here should land ahead of it.

So this is blocked on an external prerequisite rather than on effort, and the
worktree is deliberately retained: it holds the only copy of 5912f25 outside
the reflog. It is not stale scratch and should not be cleaned up while
kanon#3555 is open.