fix: filter empty preference memory items

[RerankerGuo]

Fixes #2006.

Summary

• Skip extracted preference entries whose normalized preference field is empty before creating PreferenceMemory items.
• Strip non-empty preference text before storing it in metadata.
• Add regression coverage for explicit and implicit preference extraction results that mix empty and valid entries.
• Convert touched preference-extractor log messages to parameterized logging while keeping behavior unchanged.

Validation

• python3 -m py_compile src/memos/mem_reader/read_pref_memory/process_preference_memory.py tests/mem_reader/test_preference_memory.py
• Manual line-length check for touched Python files passed.
• Attempted: python3 -m pytest tests/mem_reader/test_preference_memory.py -q; blocked because pytest is not installed in this environment.
• Attempted: make format; blocked because poetry is not installed in this environment.

[Memtensor-AI]

Automated Test Results: QUEUED

Conflict with dev-v2.0.22 was resolved by merging the base branch into this PR head. Local syntax check passed; cloud test-engine rerun has been queued.

• Run: tr-c36bee8d-bf7 on cloud test-engine 10010
• Branch tested: refs/pull/2011/head

I will post the final pass/fail result when the run completes.

[Memtensor-AI]

Automated Test Results: PASSED

Cloud test-engine rerun against dev-v2.0.22 completed successfully.

• Run: tr-c36bee8d-bf7 on cloud test-engine 10010
• memos_python_core/changed-repo-python: 2 passed, 0 failed, 0 skipped

Manual code review is still required before merge.