feat(install): persist memory learning settings

[krzkraw]

Description

Headroom already supports memory and traffic-learning controls at runtime, but persistent install manifests did not retain the learning toggle, storage mode, minimum evidence, or project-root override.

This PR persists those settings through the existing manifest, planner, state, and generated supervisor command. It also removes the duplicate legacy HEADROOM_MODE input.

This complements merged #2563 and overlaps #2346.

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to change)
• Documentation update
• Performance improvement
• Code refactoring (no functional changes)

Changes Made

• Add persistent --learn/--no-learn, --memory-storage, --min-evidence, and --memory-project-root options.
• Save and restore those values through DeploymentManifest.
• Regenerate matching runtime arguments during apply and restart.
• Leave optimization mode unset when the operator does not select one.
• Show persisted memory and learning settings in install status.

Testing

• Unit tests pass (pytest)
• Linting passes (ruff check .)
• Type checking passes (mypy headroom)
• New tests added for new functionality
• Manual testing performed

Test Output

176 passed, 2 skipped
All checks passed!
Success: no issues found in 509 source files

Real Behavior Proof

• Environment: macOS; temporary install homes and manifests plus the combined local deployment.
• Exact command / steps: Exercised learning on/off, all storage modes, custom evidence thresholds, project-root overrides, legacy manifests, restart regeneration, generated runtime arguments, and install status.
• Observed result: Settings survived manifest round-trips and produced one consistent supervisor command; omitted values retained runtime defaults.
• Not tested: Linux, Windows, VMs, or GitHub CI.

Runtime Rollout Safety

• Rollout-managed feature(s): None.
• Minimum rollout channel: N/A; this change uses stable/default behavior.
• Stable/default behavior changed: Install profiles retain learning, memory-storage, minimum-evidence, and project-root settings across apply and restart.
• Kill switch / disable path: Use --no-learn and omit --memory to disable both learning and persistent memory.
• Unsafe override required: No.
• Qualification impact: Installer manifest round-trips and generated runtime arguments for memory and learning settings.
• Rollback path: Revert this PR and regenerate affected profiles before using an older Headroom version.

Review Readiness

• I have performed a self-review
• This PR is ready for human review

Checklist

• My code follows the project's style guidelines
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• Documentation changes are not required for this installer parity change
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• I did not edit CHANGELOG.md — it is generated by release-please from my Conventional Commit PR title (a CI guard enforces this)

Screenshots (if applicable)

Not applicable.

Additional Notes

Older manifests load with backward-compatible defaults. New manifests add four optional fields; older Headroom versions may require regenerating the profile before reuse.

[github-actions]

PR governance

This PR follows the template and is marked ready for human review.

[JerrettDavis]

Reviewed the install manifest/state/CLI plumbing for persisted memory learning options. The new fields round-trip through saved manifests, old manifests get safe defaults, and install no longer writes two competing mode authorities (proxy_mode plus HEADROOM_MODE). I also merged current main into the PR branch.

Local verification after the update: uv run --extra dev python -m pytest tests/test_install/test_memory_options.py tests/test_install/test_state.py tests/test_install/test_proxy_mode_default.py tests/test_cli/test_install_cli.py -q -> 48 passed.

[JerrettDavis]

Re-reviewed current head acd054ad after the main-branch refresh. The persisted learning/storage/evidence/project-root fields still round-trip through manifest state and runtime arguments, the single mode authority remains intact, and the integration preserves current main’s Docker host-memory-path and Podman fixes. Local validation: 49 focused installer/state tests passed; touched-file Ruff check/format, targeted mypy, and diff whitespace all pass. No blocking finding.

[JerrettDavis]

Re-reviewed current head \f04986c\ after the latest rebase. The feature patch still persists learning/storage/evidence/project-root settings through manifest state and runtime arguments, retains one mode authority, preserves explicit environment precedence, and includes backward-compatible manifest migration coverage. No new blocking issue found in the refreshed patch.