Cut the antithesis density on the organisation landing page

[iderex]

profile/README.md is the organisation's public landing page, and the fleet text
survey found one thing in it: antithesis density. Seven figures over 802 words,
8.7 per 1000, which is the threshold of the tell list. Formally the page was
already clean - no em dash, no label scaffold, no emoji, no self-reference.

What changed, and what it prevents

Six rather than figures and one X, not Y in a page of 802 words read as a
mannerism, and a mannerism is what a stranger meets before they meet the product.
Four sites are rewritten:

• the opening group of three is extended to four, so each image stands for one
  row of the speed table under it, including the play-to-first-frame row that
  had no image
• sharing a core rather than a codebase says what is shared instead of what is
  not
• a failed build, not a slow one becomes has failed
• the table says so rather than letting you find out after installing becomes a
  plain sentence

Two rather than figures are kept on purpose, and they are the two where the
contrast carries the information: missing artwork as a design case against an
error case, and a preset per colour deficiency against one blanket accessible
mode. A sixth site, the security-reporting paragraph, was rewritten as well to
reach the count, and it keeps the instruction it carried.

Measured at this head

git grep -cF 'rather than' -- profile/README.md
profile/README.md:2

git grep -nP '\x{2014}'
(no line, exit 1)

git grep -nP '\x{2013}'
(no line, exit 1)

Six before, two after. The dash commands are silent before and after this change
and are quoted so the negative stays negative.

The means

Markdown, unchanged. The artefact is a rendered profile page GitHub reads from
one fixed path, so the format is forced and held to its minimum, and nothing new
enters the repository.

WHY THIS BRANCH AND NOT THE FIRST ONE

Pull request 2 carries the identical tree and one defect: its commit has no DCO
sign-off, so Verify DCO sign-off on every commit refused it. The repair is a
trailer in the message, which cannot be added to a pushed commit without
rewriting it, and I do not force-push. So the same change was committed again
with -s on this branch and pull request 2 is closed with that reason in its
body. Both heads hold the same bytes:

git diff doc/896-profile-readme-antithesis doc/896-profile-readme-antithesis-dco
(no output, exit 0)

WHAT THIS DOES NOT DISCHARGE

The fourth item on operations issue 896 asks that the text be read aloud once.
That is a person's act and nothing in this change is evidence of it, so it is
left unticked rather than reported as done. Operations issue 896 stays open for
that item alone; the three measurable items above are met at this head.

No second reader was available on this board tonight. The commands stand in place
of one.