lockers: `T& store(..., const T& cargo)` is kinda nonsensical

[xparq]

• I mean if it takes a const ref, then it shouldn't return a non-const ref to it...
  (And a templated version with perfect forwarding wouldn't solve it: by-val store shouldn't return a copy for "method chaining" :) )

• Element access is already an overload pair:

  const T& operator [](key_t key) const;
  T& operator [](key_t key);

• But... I think the problematic version should just be removed, unless there are strong use cases for it -- especially as it can't even return an error!...:

   T& store(key_t slot, const T& cargo) {
   	assert(slot != ERROR);
   	assert(!_is_free(slot));
   	return _slots[slot] = cargo; } // Support method-chaining