chore(lint): migrate to eslint-plugin-react-hooks 7.1 (set-state-in-effect) — held at ~7.0.1

[cliffhall]

Discovered while doing #1839 (the vitest/Storybook exact-pin untangle).

Regenerating clients/web/package-lock.json to untie the @vitest/browser peer knot floated eslint-plugin-react-hooks 7.0.1 → 7.1.1, which enables the new react-hooks/set-state-in-effect rule. That rule fails npm run lint on 8 pre-existing violations — none of them introduced by that PR:

File	Line
components/groups/MrtrConversation/MrtrConversation.tsx	135
components/groups/NetworkEntry/NetworkEntry.tsx	456, 469
components/groups/PromptArgumentsForm/PromptArgumentsForm.tsx	71
components/groups/ProtocolEntry/ProtocolEntry.tsx	300
components/groups/ResourceTemplatePanel/ResourceTemplatePanel.tsx	122
components/groups/ServerConfigModal/ServerConfigModal.tsx	215
components/groups/TaskCard/TaskCard.tsx	120

Six of the eight are the same shape — useEffect(() => setIsExpanded(isListExpanded), [isListExpanded]), a prop mirrored into local state so a list-wide expand/collapse control can drive a row that also has its own toggle. The remaining two are reset-on-identity-change (PromptArgumentsForm clearing completions when the prompt changes, ResourceTemplatePanel clearing variables when the template changes) and one modal-open reset (ServerConfigModal).

The hold

clients/web/package.json now declares "eslint-plugin-react-hooks": "~7.0.1" so 7.1.x cannot be picked up. This is a deliberate, temporary hold to keep a security bump from carrying an unrelated lint migration — it needs removing.

What to do

The rule is pointing at something real, so the fix is a refactor, not a suppression. React's own guidance for the dominant shape is to derive during render rather than sync in an effect — either a key to remount the row on the list-wide toggle, or the "adjust state while rendering" pattern that tracks the previous prop value. The reset-on-identity-change cases are the same story with a different trigger.

Each site changes real interaction behavior (a row's own toggle must still win over the list-wide one until the list-wide value next changes), so every one needs its existing tests kept green and, where the current tests do not already pin that precedence, a new one that does. The ≥90% per-file gate applies as usual.

Do not disable the rule to clear the errors — the whole point of taking 7.1 is the new coverage.

[cliffhall]

Superseded in practice by #1900 (the eslint 10 migration for #1838), which landed in parallel with #1839.

#1900 had to move eslint-plugin-react-hooks to 7.1.1 — not by choice: 7.0.1's peer range stops at ^9.0.0, so it ERESOLVEs against eslint 10. That forced all 8 set-state-in-effect sites listed above to be fixed, and they were — via a new useValueChange hook implementing React's documented "adjusting state during render" pattern, with the row-toggle-vs-list-toggle precedence pinned by tests (plus a new ServerConfigModal open→edit→close→reopen test).

Two follow-ups when these merge:

1. This issue should close with chore(deps): migrate to eslint 10 to clear the brace-expansion DoS advisory #1900 rather than be worked separately.
2. chore(deps): untangle the vitest/Storybook exact-pin knot to clear GHSA-p63j-vcc4-9vmv #1899 declares "eslint-plugin-react-hooks": "~7.0.1" in clients/web/package.json as the temporary hold this issue describes. That hold has to come out — if chore(deps): untangle the vitest/Storybook exact-pin knot to clear GHSA-p63j-vcc4-9vmv #1899 merges after chore(deps): migrate to eslint 10 to clear the brace-expansion DoS advisory #1900 it will pin the version back down and re-break the build against eslint 10.

Both PRs also touch clients/web/package.json and its lockfile, so whichever merges second needs a rebase regardless.

[cliffhall]

Resolved by #1900 (merged 834fd2d7), not worked separately.

#1900 had to take eslint-plugin-react-hooks@7.1.1 — 7.0.1's peer range stops at ^9.0.0 and ERESOLVEs against eslint 10 — so the temporary ~7.0.1 hold this issue tracked was removed as part of that merge, and all 8 set-state-in-effect violations were fixed via the new useValueChange hook (React's documented "adjust state during render" pattern), with the row-toggle-vs-list-toggle precedence pinned by tests.

Verified on v2/main after merge: clients/web/package.json now declares "eslint-plugin-react-hooks": "^7.1.1" with no hold remaining.