Stop the readme citing its own opening section for the other answer

[iderex]

Refs #367

Finishes: #367

What was wrong

git show fa9dd84:README.md | sed -n '265,266p'
There is nothing to run them against today. No release exists, which the section
at the top of this file says and this prints nothing for:

git show fa9dd84:README.md | sed -n '19,20p'
There are two releases, and the paragraph after the command is the one to read
before installing either:

Two hundred and forty lines apart in one file, and the paragraph carrying the
wrong answer names the paragraph carrying the right one as its authority. The
command it says prints nothing prints two rows:

gh release list --repo Flowfin/jellyfin-plugin-metadata-sync
0.1.1.0-stable  Latest  0.1.1.0-stable  2026-09-04T11:17:20Z
0.1.0.0-stable          0.1.0.0-stable  2026-09-03T10:26:03Z

The citation is what makes it hard to catch by reading. A reader who trusts the
cross-reference has no reason to scroll back and check it, and the section it
points at moved on 2026-09-05 while this paragraph did not:

git log -2 --format='%h %ad %s' --date=short fa9dd84 -- README.md
0f19252 2026-09-05 Say where this is after the second release and the catalogue entry (#88)
7c58d94 2026-09-04 Say that the rebuild after the release published nothing (#88)

The paragraph is the one about checking a downloaded archive, so an operator
holding one of the two published archives opened the section about verifying it
and was told there is nothing to verify.

What it says now

The paragraph points at Where this is today for the releases instead of
restating them, which is where the second copy came from. The correction is
marked in the text, in the shape this file already uses, so a reader who saw the
old sentence can see that it moved.

What stays negative

Nothing here claims anybody has run the checks. That half of the paragraph was
right, it is what the paragraph is for, and it is kept in the same words: the
commands are what an archive is checkable with, and this file records no run of
them. The repair removes an absence that was false and adds no assurance in its
place.

What holds it

Nothing new, and nothing is claimed. ReadmeStatementTests reads the two fenced
lists in this file and neither is in this section; no leg reads the prose around
them, which that suite's own remarks say. A release listing is not derivable
here in any case, because the suite runs with no network.

VSTEST_CONNECTION_TIMEOUT=300 DOTNET_CLI_UI_LANGUAGE=en dotnet test Jellyfin.Plugin.MetadataSync.sln --nologo
Failed: 0, Passed: 1696

per target, over both server lines, before the change and after it.

One thing about where this landed

README.md is the file the two changes above it in the log landed in, both under
#88. This is one sentence in a section neither of them touched, and it is a
separate issue rather than a widened scope on either. If it collides with work in
flight on that file, the collision is in this section and nowhere else.

Reading

No second reader looked at this. The commands above are the evidence in place of
one.