Fix configured S3 database initialization path

[FenjuFu]

What changed

• remove the hard-coded ~/.burr_server/db.sqlite3 path from the S3 initialization helper
• create the parent directory for settings.DB_PATH
• add a regression test using a custom nested database path and Tortoise configuration
• run the new optional-dependency test in the dedicated S3 CI job while allowing the default test job to skip it cleanly

Why

settings.py supports BURR_SERVER_ROOT and BURR_DB_FILENAME, but initialize_db.py independently hard-coded the default path. As a result, running the initializer could prepare a different database from the one the S3 server was configured to use.

This is intentionally scoped as a prerequisite to #524. It does not choose between Aerich migrations and schema generation as the supported production bootstrap, and it does not claim to resolve the deployment-documentation issue.

Validation

• dedicated S3 command: 22 passed
• Black 23.11.0 (--line-length 100)
• isort 5.12.0
• Flake8 6.1.0
• workflow YAML parsed successfully
• git diff --check

Contributed as a member of the iFLYTEK Astron community.

Related to #524.