Refactor implementation of column setitem

[vyasr]

This small PR reworks the behavior of ColumnBase.__setitem__ when it is provided something other than a slice as input, for instance an array. This code path requires scattering the new values into the column, which previously involved converting the Column to a Frame in order to call Frame._scatter. Since that method was only used for this one purpose, the underlying libcudf scatter implementation has been rewritten to accept and return Columns, allowing us to inline the call and also get rid of a round trip from Column to Frame and back.

[vyasr]

Although not necessarily the main purpose of this PR, it's worth noting that it does improve performance a bit, although nothing groundbreaking (a little under 10%).

Before:

(rapids) rapids@compose:~/cudf$ ipython -i -c "import cudf; import numpy as np; df = cudf.DataFrame({'a': 
Python 3.7.10 | packaged by conda-forge | (default, Feb 19 2021, 16:07:37)
Type 'copyright', 'credits' or 'license' for more information
IPython 7.26.0 -- An enhanced Interactive Python. Type '?' for help.np.arange(10000)});"

In [1]: %timeit df[df > 5000] = 10
1.24 ms ± 5.1 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

In [2]: %timeit df[df > 5000] = 10
1.25 ms ± 12.1 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

In [3]: %timeit df[df > 5000] = 10
1.24 ms ± 7.92 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

After:

(rapids) rapids@compose:~/cudf$ ipython -i -c "import cudf; import numpy as np; df = cudf.DataFrame({'a': np.arange(10000)});"
Python 3.7.10 | packaged by conda-forge | (default, Feb 19 2021, 16:07:37)
Type 'copyright', 'credits' or 'license' for more information
IPython 7.26.0 -- An enhanced Interactive Python. Type '?' for help.

In [1]: %timeit df[df > 5000] = 10
1.15 ms ± 15.6 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

In [2]: %timeit df[df > 5000] = 10
1.17 ms ± 7.39 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

In [3]: %timeit df[df > 5000] = 10
1.16 ms ± 3.7 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each)

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (branch-21.10@a153493). Click here to learn what that means.
The diff coverage is n/a.

@@               Coverage Diff               @@
##             branch-21.10    #9110   +/-   ##
===============================================
  Coverage                ?   10.76%           
===============================================
  Files                   ?      114           
  Lines                   ?    19083           
  Branches                ?        0           
===============================================
  Hits                    ?     2054           
  Misses                  ?    17029           
  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 a153493...4334941. Read the comment docs.

[marlenezw]

LGTM!!!

[shwina]

@gpucibot merge