fix(sqlalchemy-spanner): isolate test database names and extend stale cleanup cutoff

[chalmerlowe]

Problem

When running comprehensive compliance and system test suites for sqlalchemy-spanner, test sessions occasionally encountered "404 Database not found" errors. This happened because:

1. The stale database cleanup script used a 10-minute timeout threshold, which was too aggressive for longer compliance test runs that exceed 10 minutes.
2. Concurrent test runs shared a time-based naming pattern without sufficient uniqueness, leading to potential race conditions where one test run's cleanup deleted another active test run's database.

Solution

1. Updated create_test_database.py to generate session-isolated database names using a truncated timestamp and a random hex suffix (within Spanner's 30-character name limit).
2. Increased the stale database cleanup cutoff threshold to 30 minutes to accommodate long-running test suites safely without risking quota exhaustion.
3. Added active duration logging during teardown in drop_test_database.py to report how long each test database was active.

Notes for Reviewers

• This change prevents premature database dropping during concurrent test execution and provides clear test lifecycle metrics in Nox logs.

[gemini-code-assist]

Code Review

This pull request updates the test database lifecycle management for sqlalchemy-spanner. It extends the stale database cleanup threshold from 10 to 30 minutes, introduces session-isolated database IDs, and records database session metadata to track and report active duration upon teardown. The review comments point out critical issues: truncating the Unix timestamp to 9 digits changes the year to 1992, which would cause immediate deletion of newly created databases; the cleanup routine needs to handle both 10-digit and 13-digit timestamps; and the loaded JSON metadata should be type-validated as a dictionary before lookup to comply with the repository's style guide.