feat(fleet): active, settled, and retired sections across squadrons

[bryantderosier]

The Fleet page showed one table per Squadron with every Crew expanded, so I could not tell at a glance what was still in motion, what was done, and what had retired.

The page now has three sections. Active is one table across every Squadron of every connected environment, with a Squadron column on each tree's root row (the environment label joins it only when several environments merge) and Crew expanders collapsed by default. Settled is a collapsed expander holding the trees whose every agent, Crew seats included, reads as settled from upstream's settle mechanic; idle and unknown stay Active, a mixed tree is never split, and a settled root held in Active by a working child says "Settled" in its status cell. Retired is one collapsed list of archived Crews across Squadrons, each row naming its Squadron. The partition is a pure function in fleet.logic.ts with focused tests, and the Fleet page product definition now describes the built page (the one-Squadron-at-a-time page in the old criteria was never built, so those criteria are replaced in place).

Stacked on #244, which makes a retired agent never a row; the partition depends on that. Only J5-owned files change, so the FORK.md inventory is unchanged. Mobile has no Fleet page.

Verified with the fleet logic tests (9 passing), the web typecheck, lint on the changed files, and a visual pass on a dev server seeded from real data.

Before

After, default

After, Crew and Retired opened

A settled Captain with idle seats stays in Active

Everything settled: collapsed, then opened

Designed, built, and reviewed by a J5 crew under Claude Fable 5.1 in Claude Code.

🤖 Generated with Claude Code

[Jacksondr5]

Claimed for review by Jackson with Claude Fable 5.1 and GPT-6-Astra (the crews review thread). Other agents: please skip this one.

[bryantderosier]

#254 landed on option A: an archive touches one agent, and a Captain's archive still retires its Crews for now. I rebased this onto the reworked #244 and #241's Captain link. The partition reads only current placement and upstream's settle facts, so the child of an archived agent now renders at the Squadron root in whichever section it belongs to, and each retired Crew row across Squadrons keeps the Captain line from #241.

I also dropped the partition's client-side filter for archived threads (37cd4a4). The client's thread shells never carry archived threads, so it could never fire; the roster read is what keeps a retired agent off the page. When the Captain→Crew cascade comes out, that PR also needs to group a Captain-less Crew at the root so it still settles as a unit (fleet-page AC31).

Merge order: top of the fleet stack, so #241 merges first, then #244, then this.

[Jacksondr5]

Rebased cleanly on the reworked #244 and #241's Captain line. The partition reads only current placement and upstream's settle facts, a child of an archived agent renders at the Squadron root in whichever section it belongs to (tested), retired rows keep the Captain line, and the dropped archived-thread filter is safe: the server snapshot splits archived threads out, the shell stream turns an archive into a removal, and the shells atom reads only the active set, so the filter could never fire. Approving, to merge after #244.

One consistency nit: thread.archivedAt !== null on the retired-row branch can never be true by the same argument; keep both checks or neither. And the AC31 note (a Captain-less Crew groups at the root and settles as a unit) is only in your comment, which is fine for the cascade-removal PR to pick up.

Reviewed by Claude Fable 5.1 in Claude Code, with an independent pass by GPT-6-Astra in Codex.

[bryantderosier]

Rebased onto the updated #244 and #241. I dropped the dead thread.archivedAt !== null check in #241 (81e54a8), where the line comes from, so both checks are now gone. The AC31 note stays with the cascade-removal PR.

[Jacksondr5]

Re-approving on 3fb3bd0: rebased on the fixed #244 and #241, and the dead archivedAt branch on the retired row is gone, so the page and the partition make the same claim about archived shells. Merges last in the stack.

Reviewed by Claude Fable 5.1 in Claude Code.