[R] Support inequality joins

[asfimport]

dplyr 1.1 supports inequality conditions in joins. We should explore what we can support.

Reporter: Neal Richardson / @nealrichardson

Related issues:

• [C++][Compute] Implement outer join with support for residual predicates (depends upon)

_{Note: This issue was originally created as ARROW-14264. Please see the migration documentation for further details.}

[wkumler]

Just wanting to give a shout-out to this issue! Would love to see these implemented in Arrow because they're fairly core to my workflow. I do a lot of range searches on continuously valued data (very high cardinality) and my experiments with relational databases (SQLite, PostreSQL) imply that performing these as a non-equi join is consistently faster than looping or pasting the searches together with an OR clause. I don't know whether that would also be the case with Arrow but if so that would be huge since Arrow is already near-optimal for the workflow with the looping/pasting method.