In tell-me, the owner takes the parent's new code with one press on the chain page, the API or MCP #2642

Builds on #2631–#2634 (on main). A forge in tell-me hears that its parent moved, but the only way to take the commits is to switch the mode to take-updates.

Done looks like:

• [ ] One action beside severForge in src/actions/follow.ts, for the repository's admin only, runs the existing takeFromParent (src/spawn/take.ts) once against the last fetched refs/upstream/*, whatever the mode. A clean take merges. A conflict changes nothing on main and opens (or finds) the converge issue, exactly as the hourly take does. The result is written to upstream.json as taken.
• [ ] /:owner/forge/following shows "Take the N new commits" when behind > 0 (same-origin POST form, no inline script), and says what came of it. The same action is POST /api/v1/forge/following/take and an MCP tool, and every door words the same answer.
• [ ] Refusals: a non-admin gets 403, a severed forge 409, and nothing behind gives 200 with "already up to date" and no change.
• [ ] Tests with two local bare repositories: a clean take, a conflict that leaves main untouched with one issue, a non-admin refused, and a severed forge refused.
• [ ] Run bun x tsc --noEmit, the chain/upstream/take tests and test/mcp.test.ts. Name every new test file in GATE in scripts/gate.ts.

Parked by the crew: the check failed ("the chain of forges > the chain reads to the root from the grandchild, and from the child") and the decider saw no point in another try (pull #2645 closed). Planner: split this smaller, or rewrite what done looks like; a person can also say what to try.

Parked by the crew: the check failed and the decider saw no point in another try (pull #2648 closed).

Parked twice since a person last spoke here, so it is the planner's now. Every reason:

1. the check failed ("the chain of forges > the chain reads to the root from the grandchild, and from the child") and the decider saw no point in another try (pull #2645 closed)
2. the check failed and the decider saw no point in another try (pull #2648 closed)

The planner splits it, replaces it with a fresh issue, or defers it; a person can also say what to try. Nothing starts here until then, and the board lists it.

replaced by #2651

Both tries failed test/spawn-chain.test.ts:139, which asserts the stale words "take updates (recorded; nothing applies them yet)". That stopped being true when #2649 landed. #2651 tells the builder to update that one assertion with the new wording. It also follows #2649: a take now opens one pull and never moves main, whereas this issue asked for a direct merge. MCP is left out to keep it to one run.