Standardize chunking policy for bulk APIs and IN-clause lookups

Problem

Chunking is currently handled in two different ways:

1. User-facing chunksize params for write batching (for example in bulk insert APIs).
2. Method-local internal constants like CHUNK = 900 for safe IN (...) query parameter binding.

Both are valid, but behavior and naming are not standardized, which makes reviews and maintenance harder.

Why this matters

• Inconsistent patterns across methods increase review friction.
• Safe bind limits can be missed in new bulk paths.
• It is not obvious when a knob is user-facing vs internal safety.

Proposal

Define and document a single policy:

• Keep chunksize as a per-method user-facing write batching parameter.
• Keep bind-safety chunking internal (not user-facing).
• Introduce a shared internal constant/helper for IN (...) chunking (instead of ad-hoc local constants).
• Update affected methods to follow the same pattern.

Candidate code paths

• src/plexosdb/db.py (get_memberships_system and other IN (...) lookups)
• src/plexosdb/utils.py (insert_property_values membership-name lookup)
• New/ongoing bulk APIs (for example add_attributes_from_records)

Acceptance criteria

• Policy documented in developer docs/contributing notes
• Shared internal helper or constant introduced
• Existing bulk/query methods aligned
• Tests cover large inputs across chunk boundaries