chore(docs): use a builder-portable homepage check in the override template

[lesnik512]

Why

page.is_homepage is a MkDocs Page property with no equivalent in Zensical, which the org is
evaluating as a future docs builder (modern-python/.github#60). It fails silently, not loudly:
at Zensical 0.0.59 the property does not exist anywhere in the source, and an undefined name
resolves falsy in MiniJinja — so the home page would fall through to the non-home branch and emit a
wrong og:title / twitter:title with no warning.

nav.homepage is exposed by both builders — MkDocs documents it as homepage: Page | None
(https://www.mkdocs.org/dev-guide/themes/), and Zensical serializes it in
crates/zensical/src/structure/nav/view.rs — so comparing page.url against it is portable.

Part of modern-python/.github#71, which makes the same change in each repo with an
overrides/main.html.

Design

{%- set is_home = nav.homepage and page.url == nav.homepage.url %}, then not is_home where
not page.is_homepage used to be.

• The nav.homepage and guard is load-bearing: MkDocs types it Page | None. When it is None
  the expression is falsy, which matches is_homepage being False on every page of a site with
  no home page.
• The {%- left-trim is also load-bearing: without it the new tag adds a blank line to every
  rendered page.

Non-goals

Not a Zensical migration. #60 stays open and stays blocked on unrelated gaps (exclude_docs,
validation.omitted_files). This change is a no-op under MkDocs today and does not depend on that
work landing.

Verification

Full-site build diff with the pinned toolchain, main vs this branch:

uvx --with-requirements docs/requirements.txt mkdocs build --strict -d <dir>
diff -r <before> <after>

--strict passes on both, and the output is byte-identical across all 24 pages — including
index.html's <title>, og:title and twitter:title, which are what this touches.

[github-actions]

Benchmark gate

✅ gate passed

scenario	msg/s	delete/msg	WALrec/msg	WALB/msg	fpi	upd	del	dead_tup
consumer/w1/b10	926	1.000	6.80	913	242	5000	5000	10000
consumer/w1/b100	947	1.000	6.84	919	243	5000	5000	10000
consumer/w2/b10	1125	1.000	6.79	959	242	5000	5000	10000
consumer/w2/b100	1340	1.000	6.73	956	243	5000	5000	10000
consumer/w4/b10	1040	1.000	6.79	981	261	5000	5000	10000
consumer/w4/b100	1533	1.000	6.85	1033	244	5000	5000	10000
consumer/w1/b100/tfbs100	5484	0.010	6.07	1116	243	5000	5000	10000
producer/w1/b100	2215	0.000	3.04	584	0	0	0	0

Gated (fails the build): delete_calls + tuple counters (upd/del/ins) + the producer's insert_calls, exact; select_calls within +2; wal_records within a 10% band. msg/s, WAL bytes and total calls are informational (timing/FPI noise).

[lesnik512]

Force-pushed after a rebase onto main. The branch had been cut from a HEAD that already carried docs: drop the local PR template, inheriting the org default, so it was carrying a superseded draft of a change that has since merged to main in its final form. That commit is dropped; the branch is now main plus the single overrides/main.html commit. Re-verified against the new main: mkdocs build --strict on both sides, output diffed recursively, identical.