[C++][Compute] Add decimal support for arithmetic compute functions

[asfimport]

The following arithmetic functions do not support decimal:

• abs

• abs_checked

• acos

• acos_checked

• asin

• asin_checked

• atan

• ceil

• cos

• cos_checked

• floor

• is_finite (?)

• is_inf (?)

• is_nan (?)

• ln

• ln_checked

• log1p

• log1p_checked

• log2

• log2_checked

• logb (float/decimal works int/decimal does not)

• logb_checked (float/decimal works int/decimal does not)

• mode

• negate

• negate_checked

• power (float/decimal works int/decimal does not)

• power_checked (float/decimal works int/decimal does not)

• quantile

• round (ARROW-13975)

• round_to_multiple (ARROW-13975)

• sign

• sin

• sin_checked

• stddev

• tan

• tan_checked

• tdigest

• trunc

• variance

  ? - May not be applicable

  The following kernels arithmetic functions do support decimal inputs

• add

• add_checked

• atan2

• divide

• divide_checked

• equal (ARROW-13966)

• greater (ARROW-13966)

• greater_equal (ARROW-13966)

• less (ARROW-13966)

• less_equal (ARROW-13966)

• mean

• min_max

• multiply

• multiply_checked

• product

• subtract

• subtract_checked

• sum

• unique

Assignee: David Li / @lidavidm

Related issues:

• [C++] Arithmetic kernels for numeric arrays (is a child of)
• [C++] Implement aggregate compute functions for decimal datatypes (is duplicated by)
• [C++] Comparison kernel(s) for decimals (depends upon)

PRs and other links:

• GitHub Pull Request #11313

_{Note: This issue was originally created as ARROW-13130. Please see the migration documentation for further details.}

[asfimport]

Weston Pace / @westonpace:
[~yibocai#1] I extended your existing abs/negate JIRA and added info about more kernels. If you want to keep the old description let me know and I can create a new JIRA.

[asfimport]

David Li / @lidavidm:
I think we can handle most of these by casting to float64 first (if we want, we could avoid the cast and do the conversion inline instead). abs, round (and ceil, floor, etc), mode, negate, quantile, and sign should be handled directly.

[asfimport]

Yibo Cai / @cyb70289:
@westonpace, no problem.
Just noticed this jira message. I don't know why the reporter of my old jira cards becomes this strange yibocai#1, which doesn't even exists.

[asfimport]

Antoine Pitrou / @pitrou:
Issue resolved by pull request 11313
#11313