arrow

ARROW-11877: [C++] Add microbenchmark for SimplifyWithGuarantee

This adds a microbenchmark for SimplifyWithGuarantee which, especially for a large dataset, can contribute a significant amount of time to reading a dataset, as it's used to evaluate partition expressions against the filter. This was used to help investigate ARROW-11781.

Two different filters are tested: one is fully simplified, and one has had casts inserted (which will happen if you Bind() against a schema with different types).
Two different partition expressions are tested: one is fully simplified, and one compares against dictionary-encoded values (which will happen by default if you infer the schema for a Hive-partitioned, for example).

All 4 combinations are additionally tested both when the filter matches the expression and when it does not match.

Closes #9638 from lidavidm/arrow-11877

Authored-by: David Li <li.davidm96@gmail.com>
Signed-off-by: Benjamin Kietzman <bengilgit@gmail.com>