patch 8.2.0296: mixing up "long long" and __int64 may cause problems

Problem:    Mixing up "long long" and __int64 may cause problems. (John
            Marriott)
Solution:   Pass varnumber_T to vim_snprintf().  Add v:numbersize.
diff --git a/src/message.c b/src/message.c
index 9ab0eae..c2855fb 100644
--- a/src/message.c
+++ b/src/message.c
@@ -4129,7 +4129,7 @@
  * Limited support for floating point was added: 'f', 'F', 'e', 'E', 'g', 'G'.
  *
  * Length modifiers 'h' (short int) and 'l' (long int) and 'll' (long long int)
- * are supported.
+ * are supported.  NOTE: for 'll' the argument is varnumber_T or uvarnumber_T.
  *
  * The locale is not used, the string is used as a byte string.  This is only
  * relevant for double-byte encodings where the second byte may be '%'.
@@ -4371,7 +4371,7 @@
 		p++;
 		if (length_modifier == 'l' && *p == 'l')
 		{
-		    // double l = long long
+		    // double l = __int64 / varnumber_T
 		    length_modifier = 'L';
 		    p++;
 		}
@@ -4501,20 +4501,20 @@
 		    // argument is never negative)
 		    int arg_sign = 0;
 
-		    // only defined for length modifier h, or for no
-		    // length modifiers
+		    // only set for length modifier h, or for no length
+		    // modifiers
 		    int int_arg = 0;
 		    unsigned int uint_arg = 0;
 
-		    // only defined for length modifier l
+		    // only set for length modifier l
 		    long int long_arg = 0;
 		    unsigned long int ulong_arg = 0;
 
-		    // only defined for length modifier ll
+		    // only set for length modifier ll
 		    varnumber_T llong_arg = 0;
 		    uvarnumber_T ullong_arg = 0;
 
-		    // only defined for b conversion
+		    // only set for b conversion
 		    uvarnumber_T bin_arg = 0;
 
 		    // pointer argument value -only defined for p