First-class multi-host stacks (one stack → N nodes, per-host config overlay, fan-out deploy/status)

[gfargo]

First-class multi-host stacks (one stack deployed to N nodes with per-host config)

strut.conf [stacks] maps each stack to a single host. But some stacks legitimately run on every node — e.g. our buoy per-node homelab dashboard runs on compass + watch + harbor, one image (ghcr.io/gfargo/buoy), with a per-node config file (buoy.<node>.yaml) selected via BUOY_CONFIG_FILE.

Today there's no way to express "buoy deploys to these 3 hosts" — you'd deploy it 3× with --host overrides and manually pick the right per-node yaml, and strut list/drift/status only know about the one mapped host.

Ask

• Allow a stack → multiple hosts in [stacks] (e.g. buoy = compass watch harbor).
• Support a per-host config/compose overlay convention (<stack>.<host>.yaml, compose.<host>.yml, .<host>.env) auto-selected per target host.
• Make deploy/release/status/drift/health fan out across all mapped hosts (with per-host results), so "deploy buoy everywhere" and "is buoy healthy on all nodes" are one command each.

This is the natural home for fleet-wide agents (dashboards, log shippers, node exporters).