Make the load-sensitive tests wait for what they check (closes #507)

The browser test waits for each page a click opens to load, and for
Alpine.js to start on it, before reading that page; before, a read
could find an element of the page being left.

The delivery tests' drain takes what is queued without a timer. The
dispatch paths queue before they return, and the 25 ms timer could be
due by the time select looked, which then chose at random between it
and a queued task.

The test phase keeps the tests' temporary directories, and with them
their SQLite databases, on a tmpfs: waiting for the disk at each commit
was about 40% of internal/handlers' run time on a busy host.

Model: opus-5-5

script/bootstrap

@@ -11,6 +11,7 @@ ROOT="$(cd "$(dirname "$0")/.." && pwd -P)"
 
 PKGMGR=""
 SUDO=""
+APT_UPDATED=""
 
 detect_pkgmgr() {
     [ -n "$PKGMGR" ] && return 0
@@ -39,7 +40,14 @@ pkg_install() {
     detect_pkgmgr
     case "$PKGMGR" in
         nix) nix-env -iA "nixpkgs.$1" ;;
-        apt) $SUDO env DEBIAN_FRONTEND=noninteractive apt-get install -y "$2" ;;
+        apt)
+            # Package lists may be empty (fresh images); refresh once per run.
+            if [ -z "$APT_UPDATED" ]; then
+                $SUDO env DEBIAN_FRONTEND=noninteractive apt-get update
+                APT_UPDATED=1
+            fi
+            $SUDO env DEBIAN_FRONTEND=noninteractive apt-get install -y "$2"
+            ;;
         brew) brew install "$3" ;;
         apk) apk add --no-cache "$4" ;;
     esac