docs+test: skip_duplicates enforces secondary unique constraints on PostgreSQL (fixes #1049)

[dimitri-yatsenko]

Summary

Resolves #1049.

No new skip_duplicates modes are needed. PostgreSQL already enforces secondary unique constraints when skip_duplicates=True, because the generated SQL targets only the primary key:

-- PostgreSQL: ON CONFLICT (pk) DO NOTHING
-- PK duplicates silently skipped
-- secondary unique violations still raise DuplicateError

This PR documents the behavior and adds integration tests proving it works.

Backend difference (documented, not changed)

Scenario	MySQL	PostgreSQL
PK duplicate	skip	skip
Secondary unique conflict (different PK)	skip silently	raises DuplicateError

MySQL's ON DUPLICATE KEY UPDATE catches all unique key conflicts. This is a known asymmetry documented in the updated insert() docstring.

Changes

• table.py: Updated skip_duplicates docstring to explain the PostgreSQL vs MySQL difference.
• test_skip_duplicates.py: New integration test file with 6 tests (x 2 backends = 12 runs, 4 backend-specific skips):
  • test_skip_duplicates_pk_match -- PK dup silently skipped (both backends)
  • test_skip_duplicates_unique_violation_raises_on_postgres -- new PK + conflicting unique raises (PG only)
  • test_skip_duplicates_unique_on_mysql -- documents MySQL silent skip behavior
  • test_skip_duplicates_no_unique_index -- works normally without secondary indexes
  • test_skip_duplicates_composite_unique -- composite unique index enforced (PG only)
  • test_skip_duplicates_batch_mixed -- batch with PK dups + unique conflict (PG only)

Test plan

• All 12 test runs pass (8 passed, 4 skipped by backend)
• Full suite: 797 passed, 9 skipped

Generated with Claude Code