fix(oceanbase): reject legacy schema collations

[Vanillaxi]

Which issue or RFC does this PR close?

Closes #1320 .

Rationale for this change

The fresh-schema fix in #1277 makes opaque identity columns use utf8mb4_bin, but SQLAlchemy's create_all(checkfirst=True) neither alters nor rejects tables created by earlier PowerContext versions. Therefore,upgraded deployments can start with legacy utf8mb4_general_ci identity columns and still collapse case- or accent-variant scope and source IDs.

What changes are included in this PR?

• Inspect existing PowerContext VARCHAR identity columns through information_schema.COLUMNS before create_tables(),
• Reject every incompatible identity column with IncompatibleOceanBaseSchemaError.
• Add a deterministic regression test that verifies rejection happens before table creation.
• Add a real OceanBase integration test that creates the legacy pc_sources schema with utf8mb4_general_ci and verifies that the current profile rejects it.
• Document the data-only migration procedure in the English and Chinese troubleshooting guides.

Are there any user-facing changes?

Existing OceanBase/MySQL-mode deployments with incompatible identity-column collations now fail during Server startup .The error identifies the affected columns and directs operators to a backup, rebuild, and data-only restore procedure.

How was this change tested?

• uv run pytest -q tests/builtin/persistence/test_oceanbase_profile.py
tests/builtin/persistence/test_mysql_schema.py — 17 passed, 2 skipped
• The regression failed before the fix with DID NOT RAISE IncompatibleOceanBaseSchemaError; the opt-in live OceanBase case passed with POWERCONTEXT_TEST_OCEANBASE_URL.
• make check — passed
• make test — 943 passed, 10 skipped
• make docs-test — passed

AI usage statement

OpenAI Codex was used to implement and review.

[CLAassistant]

All committers have signed the CLA.