feat(memory): add hotness scoring for cold/hot memory lifecycle (#296)

[r266-tech]

Summary

Implements cold/hot memory lifecycle management as requested in #296.

Adds a hotness score that combines access frequency and recency to boost frequently-used, recently-updated contexts in search results.

Changes

New: openviking/retrieve/memory_lifecycle.py

• hotness_score(active_count, updated_at, now, half_life_days) → float [0.0, 1.0]
• Formula: sigmoid(log1p(active_count)) * exp_decay(age, half_life=7d)
• Pure function, no side effects, fully testable

Modified: openviking/retrieve/hierarchical_retriever.py

• Blends hotness score with semantic similarity in _convert_to_matched_contexts()
• final_score = (1 - alpha) * semantic_score + alpha * hotness_score
• HOTNESS_ALPHA = 0.2 (class constant, easily configurable)
• Re-sorts results after blending

New: tests/test_memory_lifecycle.py

• 13 test cases covering:
  • Boundary conditions (zero count, None updated_at, old/recent memories)
  • Monotonicity (higher count → higher score, more recent → higher score)
  • Half-life correctness
  • Blending behavior (alpha=0 preserves order, default alpha preserves semantic dominance)

Backward Compatibility

• No schema changes — leverages existing active_count and updated_at fields on Context
• No Context class modifications
• alpha=0 produces identical behavior to the current codebase
• Pure additive change: 1 new file + minimal retriever modification

Design Decisions

• sigmoid(log1p(x)) for frequency: sublinear growth prevents runaway scores from very high access counts
• Exponential decay with 7-day half-life for recency: memories lose ~50% relevance per week of inactivity
• alpha=0.2 default: mild boost that won't override strong semantic matches, but enough to surface hot memories when semantic scores are close

Closes #296

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[MaojiaSheng]

we will review soon