[experimental] Columnar query engine initial implementation

[albertlockett]

In #1342 we implemented a version of our columnar query engine (which I'm still somewhat hacking on). I thought it would be good to have an issue to track future work on this including feature gaps, performance optimizations, future ideation/experimentation, etc.

(note: list below is still a work in progress. what's documented below is not a comprehensive list of TODOs. More may be added to it as future and for some of these we can/will make dedicated issues).

Feature gaps:

• Additional signal support
  • Metrics
  • Traces
• Attribute transformations:
  • Set attribute
    • from literals(extend attributes["X"] ="Y")
    • from other expressions (things like extend attributes["event"] = <some_field> / some_func(<some_field>) / <etc.>)
  • Rename attributes (project-rename)
  • Drop attributes (project-away)
• Filtering:
  • Literal on RHS of binary expression not supported (where "WARN" == severity_text does not work)
  • Filter by body

Plan Construction & Execution: (I will add issues to expand on the bullet points in near-future)

• Columnar query engine design top level API #1417
• Mechanism for ExecutionPlans to access current batch for varying payload types #1409
• Custom ExecutionPlan impl for filtering by attributes #1410
• Perf: remove the window(row_number()) on the root batch scan
• Optimize the post-filtering applied to child batches
• mechanism to perform a "full replan" when schema changes are not compatible with current plan

Further Exploration

• supporting nested streams pipelines (iterating metrics datapoints)
• filtering by nested streams (filter metrics by datapoints)
• support fork (copy data and send down two pipelines)