Handle telegraf down gracefully; Add pushing metrics for last successful test run

Hi,

On an environment with issues with telegraf, apt-stresstest fails but not gracefully:

| apt-stresstest@machine-3:~$ /usr/local/bin/test_apt_mirrors.py -f /etc/apt-stresstest-mirrors.yaml -p bash,ubuntu-wallpapers-xenial --update-timeout 20 --download-timeout 45 -s 127.0.0.1:8094 -v
| apt_transaction_duration_seconds,test=update_1,mirror=us.archive.ubuntu.com,ip=20.62.184.178 value=0.918013
| DEBUG: Sending data to influxdb: 127.0.0.1:8094
| Traceback (most recent call last):
| File "/usr/local/bin/test_apt_mirrors.py", line 289, in <module>
| sys.exit(main())
| File "/usr/local/bin/test_apt_mirrors.py", line 267, in main
| run_tests(
| File "/usr/local/bin/test_apt_mirrors.py", line 166, in run_tests
| output_results(
| File "/usr/local/bin/test_apt_mirrors.py", line 156, in output_results
| send_to_influx(send_to, result)
| File "/usr/local/bin/test_apt_mirrors.py", line 145, in send_to_influx
| s.connect((host, port))
| ConnectionRefusedError: [Errno 111] Connection refused

I think the charm should push metrics for last successful run. We can then add prometheus alerting rules to catch those not run with say the last hour.