revert: Bollard update rolled back, closes #66

Using Bollard 0.19.x meant that network stats weren't being recorded correctly. Updated local unhealthy container to use a speed test in order to check for network IO.

docker/Dockerfile.unhealthy

@@ -2,7 +2,7 @@
 FROM alpine:latest
 
 # Install a simple utility (e.g., curl) to run as a health check
-RUN apk --no-cache add curl
+RUN apk --no-cache add curl speedtest-cli
 
 # Create a dummy file that we will use in our health check
 RUN touch /tmp/healthy
@@ -12,6 +12,6 @@ HEALTHCHECK --interval=5s --timeout=3s --retries=3 \
   CMD [ ! -f /tmp/healthy ] || exit 1
 
 # Start a basic loop that keeps the container running
-CMD ["sh", "-c", "while :; do >&2 echo 'Container is running but will be unhealthy (also printing to stderr)'; sleep 30; done"]
+CMD ["sh", "-c", "speedtest-cli; while :; do >&2 echo 'Container is running but will be unhealthy (also printing to stderr)'; sleep 30; done"]
 
 # docker build -t unhealthy-container . -f Dockerfile.unhealthy; docker run -d --name unhealthy unhealthy-container