Update custom-probability.qmd

Fix typo in definition of bivariate normal CDF. This is the standard defn of the CDF and what the `binormal_cdf` example actually computes. Verifiable in R:

```
test_binorm_cdf <- function(a,b,rho,nmc=1e5) {
    xx <- MASS::mvrnorm(nmc, c(0,0), matrix(c(1,rho,rho,1),2,2))
    mean(xx[,1]<a & xx[,2]<b)
}
```

src/stan-users-guide/custom-probability.qmd

@@ -119,7 +119,7 @@ definition of the bivariate normal cumulative distribution function
 (CDF) with location zero, unit variance, and correlation `rho`.
 That is, it computes
 $$
-\texttt{binormal}\mathtt{\_}\texttt{cdf}(z_1, z_2, \rho) = \Pr[Z_1 > z_1 \text{ and } Z_2 > z_2]
+\texttt{binormal}\mathtt{\_}\texttt{cdf}(z_1, z_2, \rho) = \Pr[Z_1 \leq z_1 \text{ and } Z_2 \leq z_2]
 $$
 where the random 2-vector $Z$ has the distribution
 $$