pedantic mode incorrectly warning about 2 priors?

@rybern Using the latest CmdStanR from github I was trying out pedantic mode (which is really cool!) and I got an unexpected warning:

stan_file <- write_stan_file(
"parameters {
  real x;
  real sigma;
}
model {
  x ~ normal(0, sigma);
  sigma ~ exponential(1);
}"
)
mod <- cmdstan_model(stan_file, stanc_options = list("warn-pedantic"=TRUE))

Compiling Stan program...
Warning:
  The parameter sigma has 2 priors.
Warning at '/var/folders/h6/14xy_35x4wd2tz542dn0qhtc0000gn/T/RtmpwDnKa2/model-e66b730e6be7.stan', line 6, column 16 to column 21:
  A normal distribution is given parameter sigma as a scale parameter
  (argument 2), but sigma was not constrained to be strictly positive.
Warning at '/var/folders/h6/14xy_35x4wd2tz542dn0qhtc0000gn/T/RtmpwDnKa2/model-e66b730e6be7.stan', line 7, column 2 to column 7:
  Parameter sigma is given a exponential distribution, which has strictly
  positive support, but sigma was not constrained to be strictly positive.

The second two warnings make sense but the first one about sigma having 2 priors isn't correct. There's only one prior on sigma. Is this a bug?

Reproduce the warning with the R and Stan snippet in the issue using stanc3 pedantic mode. Trace the pedantic warning that reports two priors and determine why it counts sigma twice. Done means the example reports only the valid support warnings while preserving the other diagnostics.