fix(tests): isolate test schemas across concurrent engine variants in integration tests

[BenjaminVijayaraj5102004]

Description & Context

Fixes schema and database collision between concurrent integration test runs (specifically test_init_project[athena_hive] and test_init_project[athena_iceberg]).

Root Cause

TestContext.__init__ previously generated self.test_id using random_id(short=True). Because self.mark was not included in the ID and random was subject to worker seed duplication in parallel xdist runs (gw0 and gw1), both Hive and Iceberg test variants generated the exact same database names (test_schema_<test_id> and sqlmesh__test_schema_<test_id>). This caused one worker to drop/modify schemas while the other was still validating tables.

Changes

1. Isolated Test IDs: Updated self.test_id in tests/core/engine_adapter/integration/__init__.py to include self.mark and a unique UUID fragment: f"{self.mark}_{uuid.uuid4().hex[:8]}".
   • self.mark ensures different table format variants (athena_hive vs athena_iceberg) never share a database name.
   • uuid.uuid4().hex[:8] guarantees worker entropy while keeping overall identifier lengths safely within the 64-character limit for engines with strict length limits (e.g. StarRocks, PostgreSQL).
2. Cross-Platform Test Fixes:
   • Converted tmp_path to .as_posix() in tests/core/engine_adapter/integration/__init__.py and conftest.py to prevent YAML scanner escape errors on Windows paths.
   • Skipped attempting to unlink :memory: catalog paths on Windows in conftest.py.

Testing

• Ran linter: uv run ruff check (passed with all checks clean).
• Ran local integration test verification: uv run pytest tests/core/engine_adapter/integration/test_integration.py -k "duckdb and test_init_project" -v (passed).

[BenjaminVijayaraj5102004]

HI maintainer can you check this PR

[cmgoffena13]

Closing: this PR was machine-generated — templated description, automated ping, no understanding shown. Read CONTRIBUTING.md if you'd like to contribute meaningfully.