test: guard paired configuration and component defaults - #1846

Purpose of the change

Configuration models and their component parameter models intentionally duplicate several scalar defaults. A change on only one side can silently make omitted values behave differently. This adds a focused regression guard for those declared pairs.

Description

• Compare 26 explicit configuration/component pairs, including retry limits, rerankers, short-term capacity, semantic limits, session cache, store settings, and seven Nebula defaults.
• Compare both exact scalar types and values through Pydantic field metadata or dataclass metadata. Required fields and default factories fail clearly instead of being mistaken for direct defaults.
• Import components lazily and construct no clients. Skip only the seven Nebula pairs when both its optional distribution and module are absent; an installed but broken SDK still fails.
• Keep partial/injected defaults and validation constraints outside this manifest. In particular, the separate RRF validation-boundary discrepancy is not changed.

Related issue

Refs #1723. This is the preventive default-parity portion, not a claim to resolve every configuration/component validation difference.

Type of change

• Project Maintenance (regression coverage; no production behavior change)

How Has This Been Tested?

Linux, Python 3.12.14, the frozen repository base environment:

• New manifest without the Nebula extra: 19 passed, 7 explicitly named skips.
• With the locked nebula5-python==5.2.2 installed: focused configuration and RRF suite 184 passed, including all 26 manifest cases.
• Changed-file Ruff, Ruff formatting, and ty checks passed; git diff --check passed.
• Controlled mutations of BM25 k1 from 1.5 to 1.75, and Neo4j False to 0, failed their exact cases. Both production files were restored byte-for-byte afterward.

Reproduce the manifest with uv run pytest packages/server/server_tests/memmachine_server/common/configuration/test_component_default_parity.py -q. Install the Nebula extra to exercise its seven entries. The complete optional-dependency and OS/Python CI matrix was not run locally.

• Unit Test

Checklist

• I have signed the commit(s) within this pull request
• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code
• I have added unit tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes (focused suites above)
• I have checked my code and corrected any misspellings

Maintainer Checklist

• Confirmed all checks passed
• Contributor has signed the commit(s)
• Reviewed the code
• Run, Tested, and Verified the change(s) work as expected

Further comments

Implementation, code review, and validation were performed with OpenAI Codex on the contributor's authorization. The human self-review box remains unchecked. The commit carries a Signed-off-by trailer and GitHub's verified signature, created through the authenticated commit API. No human review or full CI pass is claimed.