KeepEssentials rule option (as opposed to a pure drop rule)

[tdarwin]

Is your feature request related to a problem? Please describe.

Comes from a big trace problem where there's something valuable buried in a 100k+ span trace. Generally users want to drop these, but occasionally they come through via keep-error rules and the like

Describe the solution you'd like

Instead of having only drop or sample options for ?.NUM_DESCENDANTS rules (or any rule I guess) it'd be cool to have a keep_essentials option. For example

Samplers:
    __default__:
        RulesBasedSampler:
               - Name: Dealing with Errors on Giant Traces
                 Conditions:
                    - Field: error
                      Operator: exists
                    - Field: ?.NUM_DESCENDANTS
                      Operator: >=
                      Value: 15000
                 KeepEssentials: true

This would preserve the span or span event(s) where an error exists, plus their parent/ancestor spans up to the root span. This way the user would get the spans that matter from their giant trace, and be able to identify through the parent spans where in the process the damage occurred, but would not get 100k events against their honeycomb budget.

I think from their, you could also put the trace_id into the drop decision cache, so if it's a long running batch process you don't get stuff after the decision is made.

Describe alternatives you've considered

¯\_(ツ)_/¯

Additional context

No real additional concept, just liked the idea spawned by some customer questions.