fix(ci): use npm@9.x

Previously, our CI on Node.js 20 was failing silently (e.g.
the bug fixed by #1591
did make mocha exit with a non-zero exit code, but not our CI).

This is being caused by an incompatibility between npm@8.x and
Node.js 20 (which could reasonably be argued to be a Node.js bug,
the same one that the PR mentioned above accounts for). Upgrading
to npm@9.x should fix this because it includes a fix in npm
as well (npm/cli#6399).

.evergreen/install-node.sh

@@ -5,7 +5,7 @@ export BASEDIR="$PWD/.evergreen"
 if [ "$OS" == "Windows_NT" ]; then
   powershell "$(cygpath -w "$BASEDIR")"/InstallNode.ps1
   . "$BASEDIR/setup-env.sh"
-  mkdir -p "$BASEDIR/npm-8" && (cd "$BASEDIR/npm-8" && echo '{}' > package.json && npm i npm@8.x)
+  mkdir -p "$BASEDIR/npm-8" && (cd "$BASEDIR/npm-8" && echo '{}' > package.json && npm i npm@9.x)
 
   curl -sSfLO https://raw.githubusercontent.com/mongodb-js/compass/42e6142ae08be6fec944b80ff6289e6bcd11badf/.evergreen/node-gyp-bug-workaround.sh && bash node-gyp-bug-workaround.sh
 else
@@ -58,7 +58,7 @@ else
   npm cache clear --force || true # Try to work around `Cannot read property 'pickAlgorithm' of null` errors in CI
   # Started observing CI failures on RHEL 7.2 (s390x) for installing npm, all
   # related to network issues hence adding a retry with backoff here.
-  bash "$BASEDIR/retry-with-backoff.sh" npm i -g npm@8.x
+  bash "$BASEDIR/retry-with-backoff.sh" npm i -g npm@9.x
 fi
 
 . "$BASEDIR/setup-env.sh"

scripts/docker/rocky8-package.Dockerfile

@@ -7,9 +7,9 @@ RUN dnf -y install epel-release
 RUN dnf -y install python3 rpm-build dpkg-devel dpkg-dev git
 
 # Add Node.js
-RUN curl -sL https://rpm.nodesource.com/setup_16.x | bash -
+RUN curl -sL https://rpm.nodesource.com/setup_20.x | bash -
 RUN dnf install -y nodejs
-RUN npm i -g npm@8.x
+RUN npm i -g npm@9.x
 # For some reason npm@8 failed silently (!) when $HOME was
 # set to /root and consequently $HOME/.npm was not writable
 RUN mkdir -p /tmp/home