docs(extensions): add Inspect Claim Support to the extensions gallery

[avalyset]

This scorer started as #4166 (addressing #4143). The conclusion there was that
it fits better as an external package than in Inspect core, so I've packaged it
as a standalone extension (PyPI: inspect-claim-support, GitHub:
avalyset/inspect-claim-support, MIT). This PR lists it in the extensions
gallery, where external packages live — the natural follow-through from that
decision.

What it covers: a faithfulness / claim-support scorer. It assesses whether a
claimed answer is actually substantiated by the conversation transcript, not
whether it is correct in absolute terms — a grading axis Inspect doesn't have
built-in coverage for. The rubric's core property is that absence of evidence is
not treated as support: a negative claim ("I made no network calls") only scores
SUPPORTED if the transcript can actually represent that class of event;
otherwise PARTIAL or UNSUPPORTED, never SUPPORTED. A grader parse failure returns
Score.unscored() — instrument failure, kept out of the accuracy denominator —
rather than recording it as a no-answer from the model under test. The
implementation uses only Inspect's public API.

Added under the Analysis category, next to CJE — the closest neighbour, also
model-graded scorer work.

Candour: this is a new package, added here for discoverability. Happy to hold
it until it's more established if you'd rather see some traction first.