Implement a BloomFilter class that performs probabilistic membership testing using a bit array and multiple hash functions.

Your implementation must follow these rules:

• The filter uses a bit array of length size, initialized with zeros.
• The filter uses exactly hash_count hash functions for each inserted item.
• The private method _hashes(item) must produce a list of hash_count integer indices, each in the range [0, size).
• The add(item) method must set all corresponding bits for the item’s hash indices.
• The contains(item) method returns:
  • True if all bits for the item’s hash indices are set
  • False otherwise
• The filter may have false positives, but must never produce false negatives (i.e., must never return False for an item that was previously added).