Switch pid0 from bare Host to Membrane bootstrap

[lthibault]

Depends on: #29 → #28

Summary

• pid0 now bootstraps via Membrane(WetwareSession) instead of a bare Host capability
• Guest calls graft() to receive an epoch-scoped Session with Host + Executor via the WetwareSession extension
• Host side creates a static epoch channel (never advances for now) and uses build_membrane_rpc() instead of build_peer_rpc()

Changes

File	What changed
guests/pid0/build.rs	Compile all three schemas (peer, membrane, stem) with src_prefix
guests/pid0/src/lib.rs	Switch bootstrap type to Membrane(WetwareSession), call graft(), extract executor from session extension
src/cell/executor.rs	Create local watch::channel::<Epoch>(), switch to build_membrane_rpc()

Notes

• ⚠️ Heads-up: Build fixes, tests, and dev ergonomics for pid0 #31 renames Command/CommandBuilder → Cell/CellBuilder and unifies the CLI. This branch will need a rebase once Build fixes, tests, and dev ergonomics for pid0 #31 lands.
• Static epoch — real epoch advancement is a future concern
• The _epoch_tx sender is intentionally unused for now; it gets promoted to a Command field when real epoch wiring lands
• Pre-existing WIT package name conflict (cell.wit vs streams.wit) blocks full cargo check but is unrelated

Test plan

• cargo check -p pid0 --target wasm32-wasip2 passes (modulo pre-existing WIT conflict)
• cargo check --lib -p ww passes
• Full e2e test after all membrane PRs land

[lthibault]

Heads up — #31 (fix/pid0-cleanup) introduces a couple of changes that'll affect this branch when you rebase:

• Command → Cell, CommandBuilder → CellBuilder in src/cell/executor.rs and src/cell/mod.rs
• ww-cli is gone — merged into ww exec subcommand in src/cli/main.rs (client.rs deleted)
• Makefile rewritten — staged guest builds with --target-dir target, image assembly under examples/images/

Should be straightforward to resolve — the rename is mechanical and the CLI change doesn't touch any of the membrane/RPC code.