In-memory database: build records, compare-and-set and sorted scans

Context

Welcome to the platform team. Our services need a small in-memory database for session and settings data. Its public API in database/database.py is agreed and stubbed out, and the rules are in README.md.

Build levels 1 and 2 below in 30 minutes. The checks are graded one by one, so every method you get working counts even if the rest is not done.

MDB-1: Records (level 1)

Acceptance criteria

• set stores a value and get reads it back. get returns None for a missing key or field.
• compare_and_set replaces a value only when it matches expected_value, returns True or False, and never creates a field.
• compare_and_delete removes a field only when its value matches.

MDB-2: Scans (level 2)

Acceptance criteria

• scan(timestamp, key) lists every field as "field(value)", sorted by field name in plain string order.
• scan_by_prefix keeps only the fields whose name starts with the prefix.
• An unknown key, or a key whose fields were all deleted, scans as [].

Notes

• Timestamps do not matter yet; later levels use them. Keep them as the first argument.
• Run pytest tests/visible. The hidden checks call the same methods with longer sequences of calls.
• Practice task modelled on candidate reports of Meta's four-level in-memory database online assessment. Original StackYard code, not the actual question; not affiliated with Meta.