docs(implementations): smugglr http-sql profile has shipped

[ssilvius]

The directory listed the one dogfooding client the spec's credibility rests on as unshipped. A reader evaluating whether anything real speaks http-sql read that row and concluded nothing did.

Acceptance criteria mapping

"The smugglr row states the profile has shipped and names the symbol (Profile::http_sql) so a reader can verify it."

implementations.md:18 now reads "Shipped. Profile::http_sql in rafters-studio/smugglr; emits single-statement request bodies," with the implementation name itself linked to the repo. Naming the symbol is the part that matters: the old cell asserted a status and gave the reader nothing to check it against, which is how it stayed wrong for two months. Evidence for the claim is first-hand rather than relayed -- smugglr PR #132 is merged at the work source (dfde36d), confirmed at the work source rather than from a reflection that said "open, awaiting review." I also recorded what the profile actually emits, single-statement bodies, because the batch shape is the part of the spec this client does not exercise and a directory that implies full coverage overstates the dogfooding.

"No change-narrating text ("was in flight, now shipped", a dated note) -- the table describes present state."

The cell contains no reference to its own history and no date. Deliberate, and the temptation was real here since the whole point of the change is that a status flipped: a table of present state that carries its own changelog in the Notes column degrades one row at a time. Evidence: git diff main...HEAD on implementations.md is a single row replacement with no added parenthetical or footnote, and the surrounding rows are untouched.

"The status-column question is recorded but NOT fixed here; it touches every row."

Recorded in #10's body as the root cause and in the Not done section below. implementations.md:16-18 still has the three-column shape (Implementation / Form / Notes) with status living in free prose inside Notes. Unchanged in this PR.

Not done

No status column. The structural fix for the class of defect -- nothing tracks shipped-vs-in-flight, so nothing flagged this -- is a change to every row of the table plus a convention for contributors to follow in implementations.md:32's invitation. It does not belong smuggled inside a correction to one row. Recorded on #10.

Nothing about the invitation text. implementations.md:32 solicits contributors' backend as "D1 / Turso / Postgres / DO / etc". That line interacts with issue #8's dialect work (Sean ruled the spec dialect-free, so the Postgres mention is now accurate rather than aspirational) and is that issue's business, not this one's.

Closes #10