docker: fix healthcheck so the HTTP check actually runs

The condition '[ ${SKIP_CADDY} ! "true" ]' is invalid test syntax and
evaluated false in every configuration, so the curl check against /up
never ran and only php-fpm's FCGI socket was verified. A wedged Caddy or
fatally erroring app still reported healthy.

Also probe every 30s instead of 5m so failures are detected promptly, and
raise start-period to 2m since boot runs migrations and filament:optimize.

Dockerfile

@@ -93,7 +93,7 @@ COPY docker/crontab /etc/crontabs/crontab
 COPY docker/entrypoint.sh /entrypoint.sh
 COPY docker/healthcheck.sh /healthcheck.sh
 
-HEALTHCHECK --interval=5m --timeout=10s --start-period=5s --retries=3 \
+HEALTHCHECK --interval=30s --timeout=10s --start-period=2m --retries=3 \
   CMD /bin/ash /healthcheck.sh
 
 EXPOSE 80 443

Dockerfile.dev

@@ -99,7 +99,7 @@ COPY docker/crontab /etc/crontabs/crontab
 COPY docker/entrypoint.sh /entrypoint.sh
 COPY docker/healthcheck.sh /healthcheck.sh
 
-HEALTHCHECK --interval=5m --timeout=10s --start-period=5s --retries=3 \
+HEALTHCHECK --interval=30s --timeout=10s --start-period=2m --retries=3 \
   CMD /bin/ash /healthcheck.sh
 
 EXPOSE 80 443

docker/healthcheck.sh

@@ -1,9 +1,9 @@
 #!/bin/ash -e
 
-if [ ${SKIP_CADDY} ! "true" ]; then
-    curl -f http://localhost/up || exit 1
+if [ "${SKIP_CADDY:-false}" != "true" ]; then
+    curl -sf http://localhost/up || exit 1
 fi
 
 cgi-fcgi -bind -connect 127.0.0.1:9000 || exit 2
 
-exit 0
+exit 0