Fix five behavioural bugs in an LRU cache with per-entry TTL

# Task: fix the TTL cache `ttlcache.py` is an LRU cache with per-entry TTL. Users report: recently-read keys get evicted, expired keys are still counted in `len()`/`in`, entries that expire exactly at their deadline are still returned, overwriting a key does not refresh its LRU position, and `purge()` crashes. Intended behaviour: - `get` on a live key marks it most-recently-used; when the cache exceeds `maxsize` the **least**-recently-used entry is evicted. - An entry is expired when `clock() >= expires_at` (an entry with ttl=10 set at t=0 is gone at t=10). Expired entries are removed when touched by `get`, and are never visible via `len()` or `in`. - `set` on an existing key replaces the value, resets its TTL and makes it most-recently-used. - `ttl` of 0 or negative means "do not store" (and removes any existing entry for that key). - `purge()` removes all expired entries and returns how many were removed. - `hits`/`misses` count `get` outcomes. Keep the public API (constructor signature, method names). Add your own tests if you like. No network.