queue.toml realign to ALPHA levers L1/L2/L4 (rename L7→L1, L8→L2, L9→L4)

[gHashTag]

Why

The current `bin/tri-gardener/queue.toml` lanes are named after speculative ALPHA levers (L7 Muon, L8 …, L9 …). Two of those have been falsified and the third is Phase-4 backlog after trios#143 ALPHA sync at 16:38Z.

The active ALPHA lever set for the Gate-2 push is now:

Old lane	New lane	ALPHA lever
L7	L1	attention.backward correctness fix (champion-aligned)
L8	L2	JEPA-T grad-flow patch (target on top of L1)
L9	L4	h=828 → h=1024 capacity sweep (final step before Gate-2)

(L7 Muon stays in Phase-4 backlog, INV-8 boundary documented in trios#329 §7b + #330.)

What this PR will do

• Rename lanes in `bin/tri-gardener/queue.toml` and the matching `bin/tri-railway/plan21-manifest.toml` lane patterns.
• Update the `gardener_runs.lane` projection examples in `bin/tri-gardener/src/neon.rs` doc comments.
• Update the SKILL.md decision-table cheat sheet.
• Cross-link the lane meaning to the live champion record (BPB 2.1919, h=828, 2L hybrid attn) in trios#143.

Acceptance criteria

• Every L7/L8/L9 reference in `bin/tri-gardener/` is updated or explicitly preserved (with reason)
• gardener tests still 32/32 GREEN — lane names are strings, no logic depends on the literal value
• `gardener_runs` ledger does not require schema migration — `lane TEXT` accepts the new names directly
• PR body cross-links the ALPHA 16:38Z sync comment in trios#143

Refs

• Paste source: operator paste.txt 2026-04-28T03:50+07
• ALPHA champion: trios#143
• INV-8 boundary: trios#329 §7b + trios#330
• Tracker: P0: Railway MCP multi-account routing (Acc1/Acc2/Acc3) — tracker #43

Anchor: phi^2 + phi^-2 = 3 · TRINITY · NEVER STOP