fix: make Snowflake schema creation idempotent

[fengjikui]

Description

Snowflake dataset creation now uses CREATE SCHEMA IF NOT EXISTS instead of plain CREATE SCHEMA. This removes the race where two concurrent workers both pass the existence check and one fails because the other created the schema first.

The change is scoped to SnowflakeSqlClient; the generic SQL client behavior remains unchanged for other destinations.

Related Issues

• Fixes fix(sql_client) SnowflakeSqlClient.create_dataset() missing IF NOT EXISTS causes race condition with concurrent pipeline runs #3891

Additional Context

Validation run locally:

env ACTIVE_DESTINATIONS='["snowflake"]' uv run --extra snowflake pytest tests/load/snowflake/test_snowflake_table_builder.py::test_create_dataset_uses_idempotent_schema_sql -q
# 1 passed in 0.16s

uv run ruff check dlt/destinations/impl/snowflake/sql_client.py tests/load/snowflake/test_snowflake_table_builder.py
# All checks passed!

uv run ruff format --check dlt/destinations/impl/snowflake/sql_client.py tests/load/snowflake/test_snowflake_table_builder.py
# 2 files already formatted

python3 -m py_compile dlt/destinations/impl/snowflake/sql_client.py tests/load/snowflake/test_snowflake_table_builder.py