Add metrics processor for excluding based on metric label

I'm running the ops agent on a GCE instance. The instance is running an application that uses a large number of loopback devices, like `/dev/loop1`, `/dev/loop2`, etc. I do not want metrics to be stored for these loopback devices because it accounts for a substantial portion of the storage cost of my application's metrics.

Currently, all I can find is to disable the entire `agent.googleapis.com/disk/*` class of metrics, which is a blunt instrument that throws out metrics I want to track, like disk utilization on the main device `/dev/sda2`.

I'd like to be able to write a processor like this:

```yaml
metrics:
processors:
metrics_filter:
type: exclude_by_label
label: device
label_pattern: /dev/loop*
metrics_pattern:
- agent.googleapis.com/disk/*
```

Explanation: this is a new metrics filter type named `exclude_by_label`. It takes a `label` option, indicating the label to consider as part of the filter, and a `label_pattern` option, indicating pattern to be matched. If the value of the label matches the pattern, then the metric entry should be excluded. I kept `metrics_pattern` to indicate the metrics to which to apply the filter.

_duplicated from #2176 and from #2244 which got closed as inactive. do I need a bot to keep marking this as active so your bot doesn't automatically close it?_

Start by locating the metrics processor configuration entry point for metrics.processors.metrics_filter and reviewing how existing metrics filters apply metrics_pattern. Define the exclude_by_label behavior for label and label_pattern, then verify that matching loopback-device metrics are excluded while nonmatching disk metrics remain available.

Written by the indexing model from the issue text.