chore: docs/config/test accuracy sweep

[Time4Mind]

Three read-only audit agents swept the tests, documentation, and config/env surfaces for drift against the current code (post #96/#97/#98). This PR applies the corrections.

Tests — suite was already healthy

The audit found no stale tests to delete (nothing tests retired features; the suite is green because dead paths were removed, not shimmed). Two minor cleanups:

• Drop the one genuine duplicate: tests/ccbot/test_session.py::TestActiveSessions::test_set_and_clear_active_session (the create→set→clear round-trip is covered by test_session_dm.py create + delete). Kept the unique empty-initial-state guard.
• Strengthen 3 poach-guard assertions from the tautology-prone assert result is None or result.name != "AskUserQuestion" to assert result is None (verified the parser returns None for those degraded-capture panes).

Docs

• architecture.md: corrected "background sessions render their own cards" (they emit none — panel only); fixed the commands/ inventory (no /list /use /rename; added /health /help); added the modules that were missing from the inventory (logging_setup, metrics, voice_install, local_terminal, card_model, kb_mode, response_builder, context_poll [present-but-disabled], callbacks/help).
• dm-architecture.md: rewrote the slash-command block to match setMyCommands reality — published: menu/help/history/done (+ forwarded model/effort/compact/memory); hidden/typed: new/kill/stop/archive/screenshot/usage/health. /list /use /rename never existed.
• dm-multisession-spec.md: BOT_TOKEN→TELEGRAM_BOT_TOKEN, CLAUDE_BIN→CLAUDE_COMMAND, MODEL_PATH→WHISPER_MODEL_PATH; /history is published; flagged /restore-file as not-yet-implemented.
• dm-multisession-plan.md: banner marking it COMPLETED & SUPERSEDED — its file:line hotspot map targets the removed monolithic bot.py.

Config

• .env.example: removed dead SESSION_TOKEN_BUDGET_5H / MAX_5H_TOKENS / MAX_WEEKLY_TOKENS (the retired local token aggregator — never read by code); documented CCBOT_RESUME_SETTLE_TIMEOUT, CARD_EDIT_LAG, BG_STATUS_MAX, BG_STATUS_QUOTA_THRESHOLDS, LOG_LEVEL, and added a supervisor-knobs pointer.
• session.py: stale /status → Menu → Status in a comment.

Not done in this PR (flagged for a decision)

• VOICE_BACKEND=openai is unreachable: config.py coerces any non-auto/whisper/apple/off value to auto, and the settings UI allowlist excludes openai — yet transcribe.py has a live _openai_transcribe path + docs. Either add openai to the allowlists or strip the dead code/docs. A behavior decision, left out of this doc/test sweep.
• The stale "push side has no credentials" note lives in the out-of-repo global ~/.claude/CLAUDE.md, not in any tracked file — left untouched.

No behavior change (only a comment edit in src/). 560 tests pass · ruff clean · pyright 0 errors.

🤖 Generated with Claude Code