fix: protect SQLAlchemy reflection from String bytes format

[sankalpsthakur]

Summary

Fixes #920.

set_default_formats("String", "bytes") (or set_read_format("String", "bytes")) is a global read format that makes String columns decode as bytes. The core driver already protects orchestration queries via _INTERNAL_QUERY_FORMATS = {"String": "string"}, but the SQLAlchemy reflection path (cc_sqlalchemy inspector + dialect metadata methods) issued ordinary DB-API queries and was unprotected.

Under the global bytes format this caused:

1. Loud — Inspector.get_columns() / Table(..., autoload_with=...) raised TypeError: a bytes-like object is required, not 'str' on row.type.replace("\n", "").
2. Silent — get_table_names() returned [b'my_table'] instead of ['my_table'], and engine/comment comparisons like row.engine == "Dictionary" failed.

Changes

• Thread per-query query_formats through dbapi.Cursor.execute / executemany and the SQLAlchemy dialect do_execute* methods, mirroring the existing settings / execution_options plumbing.
• Apply the same _INTERNAL_QUERY_FORMATS override to all SQLAlchemy metadata queries (DESCRIBE TABLE, system.tables, SHOW TABLES / SHOW DATABASES, dictionary create SQL, etc.) via with_internal_query_formats(...).
• User SELECTs are unchanged and still honor the global bytes format.

Checklist

• Unit and integration tests covering the common scenarios were added
• A human-readable description of the changes was provided to include in CHANGELOG
• For significant changes, documentation in https://github.com/ClickHouse/clickhouse-docs was updated with further explanations or tutorials

AI disclosure

This change was drafted with AI assistance and reviewed by the author before submission.

AI/LLM disclosure

• AI coding tools (including Grok and/or Codex agent-assisted editing) were used to help draft or modify code and this PR description.
• I reviewed the complete change, understand the reasoning, and ran the reported local tests before submitting.
• This submission is original work of authorship under the project CLA / contributor terms; AI output was not pasted unreviewed.

[sankalpsthakur]

Superseded by #935, same change rebased onto current main (resolves the cursor.py conflict with #927; keeps its _leading_keyword check and best-effort metadata probe while threading query_formats through). Closing this one; please review there.