fix: revert buggy recommendation cache causing memory bloat and latency spike

[AlexanderWert]

Problem

An active latency alert fired for frontend-proxy on transaction GET /api/recommendations:

• Observed latency: 616 ms avg (last 5 min)
• Threshold: 200 ms

Root Cause

Commit 375a36eb40eb812fa6b1110585cd863e3daf8535 introduced a get_recommendations_ids() caching function in src/recommendation/recommendation_server.py that contains a critical bug:

for x in cat_response.products:
    ids_to_add.extend(cached_ids)  # BUG: appends entire cached_ids for EACH product
    ids_to_add.append(x.id)
    if len(ids_to_add) + len(cached_ids) < MAX_CACHED_IDS:
        cached_ids = cached_ids + ids_to_add

On every cache miss, ids_to_add is extended with the entire cached_ids list for each product in the catalog. This causes O(n²) memory growth, rapidly bloating cached_ids up to MAX_CACHED_IDS (2,000,000 entries). The resulting memory pressure and list operations cause severe latency degradation on the recommendation service, which propagates up through frontend → frontend-proxy.

Fix

Reverts recommendation_server.py to the original direct product catalog lookup, removing the faulty caching logic entirely.

Impact

• Affected service: recommendation → frontend → frontend-proxy
• Alert: Latency threshold rule on GET /api/recommendations (616 ms vs 200 ms threshold)
• Correlated via git.sha attribute in latency distribution analysis