refactor: standardize on absolute imports repo-wide

[tcoratger]

Summary

Convert all intra-package relative imports to absolute and enforce the convention with ruff so it cannot regress.

Why

ruff's import combining works on the literal module-path string, so a relative and an absolute import of the same module were never merged:

from lean_spec.spec.forks.lstar.containers import (...)   # absolute
from .containers import Interval                          # relative -> same module, not merged

The package mixed both styles (~50/50), which defeated auto-combining and left duplicate import blocks scattered around.

Absolute is the more standardized choice:

• PEP 8 recommends absolute imports.
• Google's Python style guide bans relative imports outright.
• The CPython standard library uses absolute throughout.

Changes

• Config: enable flake8-tidy-imports ban-relative-imports = "all" and add TID to ruff select/fixable. Relative imports are now a lint error.
• Code: ~660 relative imports across 132 files converted to absolute via ruff's TID252 autofix, then re-sorted with isort.
• Two manual fixes: ruff mis-resolved two roots because the tests/ tree uses PEP 420 namespace packages (no __init__.py chain). Corrected helpers.* → tests.lean_spec.helpers.* and gossipsub.* → tests.lean_spec.node.networking.gossipsub.*, matching the full path tests/api/conftest.py already uses.

Verification

• just check passes: ruff lint, ruff format, ty, codespell, mdformat.
• ty resolves every import path (would flag any broken one), and a sweep confirmed all remaining absolute roots map to real importable packages.
• Note: the full pytest suite was not run locally (slow); CI will exercise it.

🤖 Generated with Claude Code