[FEA] Add dot product binary op

[charlesbluca]

Closes #2448

Implements dot product for dataframes and series by using cupy.ndarray.dot on their underlying values property.

[charlesbluca]

rerun tests

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (branch-21.10@e9caed3). Click here to learn what that means.
The diff coverage is n/a.

@@               Coverage Diff               @@
##             branch-21.10    #8909   +/-   ##
===============================================
  Coverage                ?   10.77%           
===============================================
  Files                   ?      115           
  Lines                   ?    19138           
  Branches                ?        0           
===============================================
  Hits                    ?     2062           
  Misses                  ?    17076           
  Partials                ?        0           

---

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 e9caed3...f9a9200. Read the comment docs.

[marlenezw]

Changes look good to me!

[charlesbluca]

rerun tests

[vyasr]

This looks fine overall. It's unfortunate that it has to bypass the other binop machinery, but since the dot product is not strictly column by column I don't see a good alternative. Also we're dispatching to cupy rather quickly so it's not a huge issue.

[vyasr]

@charlesbluca is this ready for another round of review?

[charlesbluca]

Yup, added a special case in __array_function__ to return the proper output.

[vyasr]

Looks good! I have a couple of higher-level questions, but nothing that needs to hold this up.

1. I noticed that you added some more descriptive comments for reflected ops. Was that requested during review? Perhaps we should instead include links to the Python Data Model page, specifically the section that describes reflected binops.
2. I get that we try to match pandas everywhere, including docs, but maintaining N copies of comments like the "Among flexible wrappers..." seems bound to get out of date. Do we really need that? If we do, perhaps we should use a formatted template for those docstrings to reduce repetition (although that would be out of scope for this PR).

[charlesbluca]

Thanks @vyasr!

1. Yeah, I did that based on a review thread with @skirui-source here; thanks for that link, it definitely explains the concept in greater detail and can be used in place of my explanation.
2. Agreed, maintaining those docstrings is a lot of extra toil that can be avoided - are there docstrings in the codebase using templates currently?

[vyasr]

Regarding templates: yes, but very haphazardly and certainly not with any consistent plan across the code base. I opened #9050 to discuss this further.

Feel free to merge this whenever you've updated the documentation on reflection as you see fit.

[charlesbluca]

rerun tests

[charlesbluca]

@gpucibot merge