Fix overflow on numeric comparison (#2849)

Although #2839 fixed the overflow of exponent subtraction,
there still is possibility of overflow in the `D2U` macro.
This patch fixes the overflow in the `D2U` macro, and also
truncates the maximum digits to `DEC_MAX_DIGITS`.

src/jv.c

@@ -528,7 +528,9 @@ static decContext* tsd_dec_ctx_get(pthread_key_t *key) {
     if (key == &dec_ctx_key)
     {
       decContextDefault(ctx, DEC_INIT_BASE);
-      ctx->digits = INT32_MAX - (ctx->emax - ctx->emin - 1);
+      // make sure (Int)D2U(rhs->exponent-lhs->exponent) does not overflow
+      ctx->digits = MIN(DEC_MAX_DIGITS,
+          INT32_MAX - (DECDPUN - 1) - (ctx->emax - ctx->emin - 1));
       ctx->traps = 0; /*no errors*/
     }
     else if (key == &dec_ctx_dbl_key)

tests/jq.test

@@ -569,7 +569,7 @@ true
 true
 
 # #2825
-(1e999999999, 10e999999999) > (1e-1147483648, 0.1e-1147483648)
+(1e999999999, 10e999999999) > (1e-1147483646, 0.1e-1147483646)
 null
 true
 true