Answer cache with TTL and LRU eviction

Before the midterm many students ask Buddy the same questions. Add a cache in front of the (expensive) AI call.

Input: first line capacity ttl costPerCall (capacity ≥ 1 entries, ttl in minutes, cost in yuan). Then lines time question... with non-decreasing times in minutes. Questions are compared after trimming, converting to lower case and collapsing multiple spaces into one.

For each request:

• if the normalized question is in the cache and time - storedTime < ttl, it is a hit: print t: HIT <normalized question> and mark the entry as most recently used (its stored time does not change);
• otherwise it is a miss: print t: MISS <normalized question> (call the AI), store it with the current time (replacing an expired entry), and if the cache now holds more than capacity entries, evict the least recently used one and print evicted: <question>.

At the end print Hits: h, Misses: m, Hit rate: x% (one decimal) and AI cost: c yuan, saved: s yuan (two decimals), where cost = misses × costPerCall and saved = hits × costPerCall.

Input:

2 60 0.05
0 What is polymorphism?
5 what is  POLYMORPHISM?
10 What is a class?
12 What is an interface?
15 What is polymorphism?
72 What is an interface?

Output:

0: MISS what is polymorphism?
5: HIT what is polymorphism?
10: MISS what is a class?
12: MISS what is an interface?
  evicted: what is polymorphism?
15: MISS what is polymorphism?
  evicted: what is a class?
72: MISS what is an interface?
Hits: 1, Misses: 5, Hit rate: 16.7%
AI cost: 0.25 yuan, saved: 0.05 yuan