Add LogicalPlan::Distinct

[mirober]

Which issue does this PR close?

Closes #2573.

Rationale for this change

What changes are included in this PR?

Adds a new type of logical plan for DISTINCT operations, which is converted into an aggregation during physical planning.

Something I am unsure of: all of the LogicalPlan structs contain Arc<LogicalPlan> where they refer to other plans, except Union, which has a Vec<LogicalPlan>. I think this caused a couple of extra LogicalPlan clones due to needing an owned LogicalPlan to put into a union. Should Union contain Vec<Arc<LogicalPlan>> instead?

Are there any user-facing changes?

[codecov-commenter]

Codecov Report

Merging #2792 (bc4193e) into master (3da59e8) will increase coverage by 0.18%.
The diff coverage is 84.21%.

@@            Coverage Diff             @@
##           master    #2792      +/-   ##
==========================================
+ Coverage   84.94%   85.13%   +0.18%     
==========================================
  Files         272      273       +1     
  Lines       48221    48293      +72     
==========================================
+ Hits        40963    41113     +150     
+ Misses       7258     7180      -78     

Impacted Files	Coverage Δ
...tafusion/optimizer/src/common_subexpr_eliminate.rs	94.11% <ø> (ø)
datafusion/optimizer/src/projection_push_down.rs	98.08% <ø> (ø)
datafusion/proto/src/logical_plan.rs	17.63% <22.22%> (+0.01%)	⬆️
datafusion/expr/src/logical_plan/plan.rs	74.21% <83.33%> (+0.30%)	⬆️
datafusion/expr/src/logical_plan/builder.rs	89.73% <96.29%> (+0.22%)	⬆️
datafusion/core/src/physical_plan/planner.rs	81.02% <100.00%> (+0.15%)	⬆️
datafusion/expr/src/utils.rs	90.73% <100.00%> (+0.05%)	⬆️
datafusion/sql/src/planner.rs	81.34% <100.00%> (-0.02%)	⬇️
datafusion/core/src/datasource/listing/table.rs	89.13% <0.00%> (-1.23%)	⬇️
datafusion/core/src/config.rs	90.32% <0.00%> (-0.79%)	⬇️
... and 12 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 3da59e8...bc4193e. Read the comment docs.

[alamb]

cc @andygrove

[andygrove]

Thanks @mrob95. I plan on reviewing this later this week.

[andygrove]

Thanks @mrob95 this looks great. I think you raise a good point about Union containing Vec<LogicalPlan> rather than Vec<Arc<LogicalPlan>> but I think that should be fixed as a separate issue/PR.

[andygrove]

@mrob95 There is a conflict that needs resolving. I can merge once that is resolved.

[mirober]

Thanks, should be fixed now 🙂

[alamb]

Filed #2816 to track the Union inconstency