Fix -Wimplicit-const-int-float-conversion warnings

Compare against 0x1p63 instead of INT64_MAX. Converting INT64_MAX to
double rounds it up to INT64_MAX+1.

It made code like `if (d <= INT64_MAX) v = (int64_t)d;` behave subtly
wrong when `d >= 0x1p63` because then `v = (int64_t)d` wraps around to
a negative value.

quickjs.c

@@ -10738,7 +10738,7 @@ static int JS_ToInt64SatFree(JSContext *ctx, int64_t *pres, JSValue val)
             } else {
                 if (d < INT64_MIN)
                     *pres = INT64_MIN;
-                else if (d > INT64_MAX)
+                else if (d >= 0x1p63)
                     *pres = INT64_MAX;
                 else
                     *pres = (int64_t)d;
@@ -53844,7 +53844,7 @@ static JSValue js_atomics_wait(JSContext *ctx,
     }
     if (JS_ToFloat64(ctx, &d, argv[3]))
         return JS_EXCEPTION;
-    if (isnan(d) || d > INT64_MAX)
+    if (isnan(d) || d >= 0x1p63)
         timeout = INT64_MAX;
     else if (d < 0)
         timeout = 0;