[AMCC-168]Allow configuring metric names, tag names and values

[StephenWakely]

What does this PR do?

This allows lading to be configured so the dogstatsd generator will output metric names, tag names and values chosen from a preconfigured list.

Configuration is picked up from the file as:

        dogstatsd:
          metric_names:
            - thing
            - thung
          tag_names:
            - thing.thong
            - ning.nung
          tag_values:
            - thingthing
            - thangthang

Motivation

I'm setting up some benchmarks to observe the performance of tag aggregation - a new feature in the agent where you can configure given tags to be removed from metrics with given names. Given the performance will only be impacted when it processes a metric and tag combination that it has been configured to filter, we need to be able to ensure that lading can generate known metrics and tags.

Related issues

The PR I would like to exercise: DataDog/datadog-agent#44166

Additional Notes

The Payload Fingerprint verification job is failing. I'm not 100% sure what that is about, do I just need to update the fingerprint?

[StephenWakely]

This is slightly questionable. With multiple string pools flying around now, there is a possibility that you could get a handle from one string pool and then try to fetch the string from another.

I've made it so it panics if you fetch a handle from a RandomStringPool and then try to use it on a StringListPool, but that still won't catch if you are just playing with different StringListPoolss.

I'm all ears if there is a better way around this.

[blt]

I've been thinking for a while this part of the program needs to be rationalized. I think this is a good risk hedge and is indicative of the need to rethink all this -- outside of this line of work.

[StephenWakely]

I've added enum_dispatch to handle the different types of pools we can have. Hope that is Ok. I tried to make everything generic over the different possible set of pools, but that resulted in a huge explosion of generics pretty much everywhere, which was not pleasant. Can try using dyn traits instead if you think that would be better.

[blt]

I find this a pragmatic choice for this PR.

[blt]

I've been thinking for a while this part of the program needs to be rationalized. I think this is a good risk hedge and is indicative of the need to rethink all this -- outside of this line of work.

[blt]

I find this a pragmatic choice for this PR.