Add a general purpose processor based on the Starlark language

[danielnelson]

Feature Request

The Starlark language is a dialect of python designed to be embedded into a larger application. The Go implementation that allows embedding the language using pure Go code and with few dependencies. It is used primarily in the Bazel build system and is well maintained.

Proposal:

Create a processor plugin that allows applying a starlark function to Telegraf metrics. Expose and document an interface for reading and modifying the metric name, tags, fields, time.

Desired behavior:

Example configuration, this script does not modify the metric but prints the values. Calling print in the processor will be connected to the Telegraf log:

[[processors.starlark]]
  source = """
def apply(metric):
    print(metric.name)
    for tag in metric.tags:
        key, value = tag
        print('tag: %s=%s' % (key, value))
    for field in metric.fields:
        key, value = field
        print('field: %s=%s' % (key, value))
    print('time:', metric.time)
    return metric
"""

Use case:

This would be a general purpose plugin which will be helpful with specialized processing and extending Telegraf without requiring new plugins. I'll be adding some additional use cases as they come up and are collected from the backlog:

• Performing simple mathematical operations such as the division of two fields.

[danielnelson]

Would also implement #3709