Refactor: extract `commands/install.py` helpers into `apm_cli/install/` package

[sergio-sisternes-epam]

Context

src/apm_cli/commands/install.py is at ~1700 LOC and is guarded by an architecture-invariant test (tests/unit/install/test_architecture_invariants.py::test_install_py_under_legacy_budget) with a legacy budget of 1680 LOC.

PR #803 (Codex project-scoped MCP, fixes #502) needs to thread scope context (project_root, user_scope, explicit_target) through install orchestration, which pushes the file to 1701 LOC and trips the invariant. Rather than block the Codex PR on a large refactor, we're raising the budget in-line there and tracking the extraction here.

Why this matters

The invariant test message says it explicitly:

Do NOT trim cosmetically -- engage the python-architecture skill (.github/skills/python-architecture/SKILL.md) and propose an extraction into apm_cli/install/.

install.py is a well-known God-module candidate. Every cross-cutting change (scope, targets, MCP, hooks, lockfile) adds pressure on it, and silent budget bumps would defeat the guard.

Proposed direction

Extract cohesive helpers into a new apm_cli/install/ package, keeping commands/install.py as the thin CLI entrypoint. Likely candidates:

• Scope/target resolution helpers (project_root, user_scope, explicit_target plumbing).
• MCP integration orchestration (dep collection, conflict detection, stale cleanup glue).
• Dependency/runtime detection and filtering.
• Post-install reporting / diagnostics assembly.

Engage the python-architecture skill before cutting module boundaries -- the goal is cohesion by responsibility, not file-size juggling.

Acceptance

• commands/install.py back under a tight budget (target: 1200 LOC or lower).
• New apm_cli/install/ package with clear module boundaries and its own tests.
• Architecture invariant updated to guard the new structure (not just the legacy LOC ceiling).
• No behaviour change for end users -- public CLI surface and flags unchanged.

Related

• PR feat(codex): add project-scoped MCP and user target support #803 -- project-scoped Codex MCP (trips the current budget; bumps it in-line with a reference to this issue).
• Fixes Support Codex CLI as integration target #502 -- the original Codex user-scope request that drove PR feat(codex): add project-scoped MCP and user target support #803.