Change calcGCD() to use the int64_t type

If `a` is of the int32_t type and equal to INT_MIN, then `-a` cannot be
represented in int32_t (assuming two's complement). To allow us to
negate INT_MIN, use the int64_t type instead.

Fix https://crbug.com/oss-fuzz/48135.

src/avif.c

diff --git a/src/avif.c b/src/avif.c
index 739066c..5ee6ea6 100644
--- a/src/avif.c
+++ b/src/avif.c
@@ -456,7 +456,9 @@
     return f;
 }
 
-static int32_t calcGCD(int32_t a, int32_t b)
+// |a| and |b| hold int32_t values. The int64_t type is used so that we can negate INT_MIN without
+// overflowing int32_t.
+static int64_t calcGCD(int64_t a, int64_t b)
 {
     if (a < 0) {
         a *= -1;
@@ -466,7 +468,7 @@
     }
     while (a > 0) {
         if (a < b) {
-            int32_t t = a;
+            int64_t t = a;
             a = b;
             b = t;
         }
@@ -477,7 +479,7 @@
 
 static void clapFractionSimplify(clapFraction * f)
 {
-    int32_t gcd = calcGCD(f->n, f->d);
+    int64_t gcd = calcGCD(f->n, f->d);
     if (gcd > 1) {
         f->n /= gcd;
         f->d /= gcd;