(Analysis by Brian Dean)

This problem can be solved using a recursive "depth first" search of the space of all possible (location, boot index) states we can reach, starting from (location 0, boot index 0).

There are $O(NB)$ states, and each time we visit a state we may try visiting $O(N + B)$ neighboring states, so the total running time will be at worst $O(N^2B + NB^2)$, which is comfortable since $N$ and $B$ are both at most 250.

Code for my recursive search approach is below:

#include <iostream>
#include <fstream>
#include <algorithm>
using namespace std;
     
int N, B, F[250], S[250], D[250], best = 9999;
bool beenthere[250][250];  // set of (location, boot) states we can reach

// recursively search through all reachable states     
void visit(int i, int b)
{
   // don't re-visit already-visited states
   if (beenthere[i][b]) return;   
   beenthere[i][b] = true;     

   // reached the end?  keep track of smallest boot # we can achieve
   if (i==N-1) { best = min(best, b); return; }

   // try all possible steps forward
   for (int i2=i+1; i2<N && i2-i<=D[b]; i2++) 
      if (F[i2]<=S[b]) visit(i2,b);

   // try all possible changes of boots
   for (int b2=b+1; b2<B; b2++) 
      if (F[i]<=S[b2]) visit(i,b2);
}
     
int main(void)
{
   ifstream fin ("snowboots.in");
   ofstream fout ("snowboots.out");
   
   fin >> N >> B;
   for (int i=0; i<N; i++) fin >> F[i];
   for (int i=0; i<B; i++) fin >> S[i] >> D[i];
   
   visit(0, 0);
   fout << best << "\n";
  
   return 0;
}

Although perhaps not the intended way of thinking about a problem in the silver division, this problem can also be solved with more of a "dynamic programming" outlook, where we loop over all possible (location, boot) states Y and compute whether we can reach that state based on prior states X that can transition to Y. This approach is actually quite similar to the recursive search, just it loops over all the states and computes their reachability one by one, instead of recursively fanning out to discover the set of reachable states. Worst-case running time would be the same as with the prior approach.

#include <iostream>
#include <fstream>
#include <algorithm>
using namespace std;
     
int N, B, F[250], S[250], D[250], best = 9999;
bool beenthere[250][250];  // set of (location, boot) states we can reach

int solve(void)
{
   for (int b=0; b<B; b++) 
      for (int i=0; i<N; i++) {

         // Compute whether state (i, b) is reachable...
	
         if (F[i] > S[b]) { beenthere[i][b] = false; continue; } // invalid state; snow too deep

         if (i==0 && b==0) beenthere[i][b] = true;  // initial state

         // can we reach this state via forward step from some prior state?
         for (int i2=0; i2<i; i2++) 
            if (beenthere[i2][b] && i2+D[b]>=i) beenthere[i][b] = true;

         // can we reach this state via change of boot from some prior state?
         for (int b2=0; b2<b; b2++) 
            if (beenthere[i][b2]) beenthere[i][b] = true;

         // have we reached the end?  return smallest boot # we can achieve
         if (i==N-1 && beenthere[i][b]) return b;
      }
}
     
int main(void)
{
   ifstream fin ("snowboots.in");
   ofstream fout ("snowboots.out");
   
   fin >> N >> B;
   for (int i=0; i<N; i++) fin >> F[i];
   for (int i=0; i<B; i++) fin >> S[i] >> D[i];

   fout << solve() << "\n"; 
   return 0;
}

Note that there are ways to optimize the code above (e.g., by breaking out of loops when we've determined that beenthere can be set to true); however, this isn't necessary to solve the problem in time.