Preserve metric type when using filters in output plugins

[roxit]

Resolves #4501

Required for all PRs:

• Signed CLA.
• Associated README.md updated.
• Has appropriate unit tests.

Steps to reproduce:

Using telegraf 1.7.2 with the following config to export kube-scheduler metrics:

[[outputs.prometheus_client]]
  listen = "0.0.0.0:9273"
  expiration_interval = "60s"
  [outputs.prometheus_client.tagpass]
    _OUTPUT = ["prometheus"]

[[inputs.prometheus]]
  urls = ["http://127.0.0.1:10251/metrics"]
  [inputs.prometheus.tags]
    _OUTPUT = "prometheus"

Histograms in http://127.0.0.1:9273/metrics look like:

...
# HELP scheduler_scheduling_algorithm_latency_microseconds_16000 Telegraf collected metric
# TYPE scheduler_scheduling_algorithm_latency_microseconds_16000 untyped
scheduler_scheduling_algorithm_latency_microseconds_16000{_OUTPUT="prometheus",host="xxx",url="http://127.0.0.1:10251/metrics"} 0
# HELP scheduler_scheduling_algorithm_latency_microseconds_1_024e_06 Telegraf collected metric
# TYPE scheduler_scheduling_algorithm_latency_microseconds_1_024e_06 untyped
scheduler_scheduling_algorithm_latency_microseconds_1_024e_06{_OUTPUT="prometheus",host="xxx",url="http://127.0.0.1:10251/metrics"} 0
...

With the tagpass filter removed from outputs.prometheus_client, the output preserved the metric type:

...
# HELP scheduler_scheduling_algorithm_latency_microseconds Telegraf collected metric
# TYPE scheduler_scheduling_algorithm_latency_microseconds histogram
...
scheduler_scheduling_algorithm_latency_microseconds_bucket{_OUTPUT="prometheus",host="xxx",url="http://127.0.0.1:10251/metrics",le="16000"} 0
scheduler_scheduling_algorithm_latency_microseconds_bucket{_OUTPUT="prometheus",host="xxx",url="http://127.0.0.1:10251/metrics",le="1.024e+06"} 0
...

Fix

The fix is simple, just pass the metric type when recreating the filtered metric in RunningOutput.
I tried to add a test for it but it seems testutil.TestMetric needs to updated to support metric type first, which seems too big to be included in this commit.