fix(persistence): compile dialect-safe boolean defaults

[justonemorenight]

Summary

• compile Python boolean defaults as typed SQLAlchemy literals for the active database dialect
• emit 0/1 for SQLite and false/true for PostgreSQL during additive column reconciliation
• cover both boolean values on both supported dialects

This fixes legacy PostgreSQL upgrades for every model Boolean backed only by a Python default, including both conversations.pinned and coverage_files.mapping_partial.

Related Issues

Fixes #2059

Test Plan

• uv run pytest -q tests/unit/persistence/test_schema_reconciliation.py (13 passed)
• uv run ruff check packages/core/src/repowise/core/persistence/database.py tests/unit/persistence/test_schema_reconciliation.py
• git diff --check

Risk

Repowise change risk: 18th percentile (Below typical / low review priority). Numeric and string defaults retain their existing rendering; only Python bool defaults use the dialect compiler.

[RaghavChamadiya]

Right fix in the right place. Compiling the typed literal against the dialect is what _add_column_ddl should have been doing from the start: bool is not int in PostgreSQL, and DEFAULT 0 on a BOOLEAN column is a syntax error there, so the legacy-column path could not add a boolean column to a Postgres index at all. The parametrized test pinning 0/1 on SQLite and false/true on Postgres is the assertion I would have asked for.

Non-blocking: the diff carries two formatting-only hunks (the blank line at database.py:46 and the f-string reflow in the _run lambda) that are not part of the fix. Not worth another round trip, but this repo does not run ruff format over the tree, so please keep future diffs to the change itself.