SNOW-1489371: Implement GroupBy.value_counts

[sfc-gh-joshi]

1. Which Jira issue is this PR addressing? Make sure that there is an accompanying issue to your PR.

   Fixes SNOW-1489371

2. Fill out the following pre-review checklist:

   • I am adding a new automated test(s) to verify correctness of my new code
     • If this test skips Local Testing mode, I'm requesting review from @snowflakedb/local-testing
   • I am adding new logging messages
   • I am adding a new telemetry message
   • I am adding new credentials
   • I am adding a new dependency
   • If this is a new feature/behavior, I'm adding the Local Testing parity changes.

3. Please describe how your code solves the related issue.

This PR adds support for GroupBy.value_counts, accepting all parameters except bin, which we do not support for DataFrame/Series.value_counts.

Upstream modin defaults to pandas for both DataFrameGroupBy/SeriesGroupBy.value_counts, so some of these changes should be eventually upstreamed.

pandas has different behavior than what might be expected from documentation; this PR tries to align with existing behavior as much as possible. This is documented in this pandas issue: pandas-dev/pandas#59307

1. When normalize=True, pandas sorts by the pre-normalization counts, leading to counterintuitive results. This only matters when groupby is called with sort=False and value_counts with sort=True. See test cases for an example.
2. pandas does not always respect the original order of data, depending on the configuration of sort flags in groupby and the value_counts call itself. The behaviors are as follows (copied from query compiler comments):

# pandas currently provides the following behaviors based on the different sort flags.
# These behaviors are not entirely consistent with documentation; see this issue for discussion:
# https://github.com/pandas-dev/pandas/issues/59307
#
# Example data (using pandas 2.2.1 behavior):
# >>> df = pd.DataFrame({"X": ["B", "A", "A", "B", "B", "B"], "Y": [4, 1, 3, -2, -1, -1]})
#
# 1. groupby(sort=True).value_counts(sort=True)
#   Sort on non-grouping columns, then sort on frequencies, then sort on grouping columns.
# >>> df.groupby("X", sort=True).value_counts(sort=True)
# X  Y
# A   1    1
#     3    1
# B  -1    2
#    -2    1
#     4    1
# Name: count, dtype: int64
#
# 2. groupby(sort=True).value_counts(sort=False)
#   Sort on non-grouping columns, then sort on grouping columns.
# >>> df.groupby("X", sort=True).value_counts(sort=True)
# X  Y
# X  Y
# A   1    1
#     3    1
# B  -2    1
#    -1    2
#     4    1
# Name: count, dtype: int64
#
# 3. groupby(sort=False).value_counts(sort=True)
#   Sort on frequencies.
# >>> df.groupby("X", sort=False).value_counts(sort=True)
# X  Y
# B  -1    2
#     4    1
# A   1    1
#     3    1
# B  -2    1
# Name: count, dtype: int64
#
# 4. groupby(sort=False).value_counts(sort=False)
#   Sort on nothing (entries match the order of the original frame).
# X  Y
# B   4    1
# A   1    1
#     3    1
# B  -2    1
#    -1    2
# Name: count, dtype: int64
#
# Lastly, when `normalize` is set with groupby(sort=False).value_counts(sort=True, normalize=True),
# pandas will sort by the pre-normalization counts rather than the resulting proportions. As this
# is an uncommon edge case, we cannot handle this using existing QC methods efficiently, so we just
# update our testing code to account for this.
# See comment 

[sfc-gh-vbudati]

Great work, Jonathan! This seems like a monstrous task, thank you for having such detailed documentation about it. I left a couple comments and I think it's good to merge after.