[C++] Implement serial version of sort computational kernel

[asfimport]

Implement serial version of sort computational kernel.

Reporter: Areg Melik-Adamyan / @aregm

Related issues:

• [C++] Implement non-materializing sort kernels (relates to)
• [C++] Implement kernels for in-place sorting of fixed-width contiguous arrays (relates to)

_{Note: This issue was originally created as ARROW-4631. Please see the migration documentation for further details.}

[asfimport]

Wes McKinney / @wesm:
I linked a couple of related issues

[asfimport]

Areg Melik-Adamyan / @aregm:
Can you please add Artem Alekseev @fexolm (artem.alekseev@intel.com)  to Jira to assign him this issue?

[asfimport]

Kirill Lykov / @KirillLykov:
Do we still need this feature? It sounds like an interesting exercise 

[asfimport]

Antoine Pitrou / @pitrou:
Unless I'm mistaken, this is now done through the sort_indices compute function.