B'This is the harder version of the problem. In this version, 1 <= n, m <= 2 cdot10^5 . You can hack this problem if you locked it. But you can hack the previous problem only if you locked both problems. You are given a sequence of integers a=[a_1,a_2, ... ,a_n] of length n . Its subsequence is obtained by removing zero or more elements from the sequence a (they do not necessarily go consecutively). For example, for the sequence a=[11,20,11,33,11,20,11] : Suppose that an additional non-negative integer k ( 1 <= k <= n ) is given, then the subsequence is called optimal if: Recall that the sequence b=[b_1, b_2, ... , b_k] is lexicographically smaller than the sequence c=[c_1, c_2, ... , c_k] if the first element (from the left) in which they differ less in the sequence b than in c . Formally: there exists t ( 1 <= t <= k ) such that b_1=c_1 , b_2=c_2 , ..., b_{t-1}=c_{t-1} and at the same time b_t<c_t . For example: You are given a sequence of a=[a_1,a_2, ... ,a_n] and m requests, each consisting of two numbers k_j and pos_j ( 1 <= k <= n , 1 <= pos_j <= k_j ). For each query, print the value that is in the index pos_j of the optimal subsequence of the given sequence a for k=k_j . For example, if n=4 , a=[10,20,30,20] , k_j=2 , then the optimal subsequence is [20,30] -- it is the minimum lexicographically among all subsequences of length 2 with the maximum total sum of items. Thus, the answer to the request k_j=2 , pos_j=1 is the number 20 , and the answer to the request k_j=2 , pos_j=2 is the number 30 . The first line contains an integer n ( 1 <= n <= 2 cdot10^5 ) -- the length of the sequence a . The second line contains elements of the sequence a : integer numbers a_1, a_2, ... , a_n ( 1 <= a_i <= 10^9 ). The th'...