Implement a capacity-bounded LRU (Least Recently Used) cache.

cache = LRUCache(capacity=2)
cache.put("A", 10)
cache.put("B", 20)
cache.get("A")      # 10   ← A is refreshed as recently used
cache.put("C", 30)  #      ← over capacity → B, the least recently used, is evicted
cache.get("B")      # -1

Core rules

• get also refreshes recency. Merely reading a key makes it "just used". Miss this and half the tests fail.
• A put on an existing key overwrites the value and refreshes its recency too.
• When capacity is exceeded, discard the single least recently used key.
• A get on a missing key returns -1 (not an exception).

Level structure

LRUCache(capacity: int)
get(key) -> value or -1
put(key, value) -> None

• capacity is an integer of at least 1.
• get returns the value and refreshes that key's recency.
• put inserts a new key or overwrites an existing one; either way it refreshes recency.
• On overflow, discard the least recently used key.