Story: Environment provisioning overhaul: lightweight environments, GCP-style naming, bootstrap independence

Two types controlled by the --type flag at compass env provision time, stored as ORES_PROVISION_TYPE in .env.

• light: no C++ toolchain, no vcpkg, no NATS, no Postgres. Supports agile docs, compass, and emacs-driven build targets (settings, skills, manual, site). Minimum bootstrap: git, emacs, python, compass venv.
• full: complete stack — C++, vcpkg, NATS, Postgres, Qt client. Superset of light.

The genesis environment is always full. New worktrees default to full unless --type light is given.

Replace ad-hoc local1=/=remote labels with descriptive adjective-noun pairs (e.g. festive-hawking, clever-turing). The name is stored as ORES_ENV_NAME in .env and used as the source of truth for all derived identifiers:

name_underscored replaces hyphens with underscores. name_dotted replaces hyphens with dots. This fixes the latent bug in env_init.py where db_name = f"ores_dev_{label}" could embed a hyphen, producing an unquoted PostgreSQL identifier that is syntactically invalid.

ORES_CHECKOUT_LABEL is kept as a deprecated alias written to .env alongside ORES_ENV_NAME during the transition; downstream code migrates to ORES_ENV_NAME over time.

The genesis environment is the canonical first clone of the repo. Convention:

• Path: ~/Development/OreStudio/OreStudio.genesis
• Type: always full
• Branch: typically main; the parent in git worktree list for all other worktrees
• Name: genesis (not an adjective-noun pair — it is the root, not a worktree)

compass bearings checks for the genesis path and emits a warning when it cannot be found, so users on a fresh machine are alerted immediately.

All new worktrees are created with git worktree add from the genesis clone (by compass env create). Existing worktrees (local1, remote) are not renamed automatically; the user migrates on their own timeline, or leaves legacy names in place — the new scheme is additive.

The hardcoded dict {​"remote": 50000, "local1": 51000, ...} and the trailing-digit NATS port derivation are replaced by values written into .env at compass env create time.

Assignment strategy at create time:

1. Scan all existing worktrees, read ORES_BASE_PORT from each .env.
2. Assign the lowest free slot: base ports 50000, 51000, 52000, ... (step 1000); NATS ports 42221, 42222, ...
3. Write ORES_BASE_PORT and ORES_NATS_PORT into the new .env before handing off to env configure.

compass env configure reads ORES_BASE_PORT and ORES_NATS_PORT from existing (the existing .env values) if already set, so re-running configure on an existing env never reassigns ports.

Port layout within a block (base_port = ORES_BASE_PORT):

• HTTP port: base_port + 0
• WT port: base_port + 2 (release: base_port + 1 / base_port + 3)
• Site port: base_port + 4

compass env create <name> [--type light|full]   NEW: provision worktree + configure .env
compass env configure [--preset P] [-y]         RENAMED from: compass env init
compass env init [--preset P] [-y]              ALIAS for: compass env configure (backwards compat)
compass env list [--show-secrets]               unchanged
compass env diff                                unchanged
compass env version [new <desc>]                unchanged

Rationale for the rename: compass env init sounds like "initialise a new environment" (worktree creation) but actually only generates the .env config file. Renaming it to configure makes the distinction clear: create provisions the worktree; configure (re)generates its credentials.