Feat: add skew and kurtosis aggregators.

[JovanVeljanoski]

Notes:

• This PR has been branched off ✨ last aggregor, and first and last now can have mixed types #1848 and thus requires that branch to be merged first since we are using the refactoring done there
• Closes [FEATURE-REQUEST] prod / skew / kurtosis aggregations #1215

Checklist:

• implement 'df.skew' and vaex.agg.skew
• implement 'df.kurtosis and vaex.agg.kurtosis
• Unit tests for everything
• Code review

[maartenbreddels]

nice work!

[maartenbreddels]

A different solution which would be shorter would be something like this:

@@ -574,7 +576,17 @@ def var(expression, ddof=0, selection=None, edges=False):
 @register
 def skew(expression, selection=None, edges=False):
     '''Create a skew aggregation.'''
-    return AggregatorDescriptorSkew('skew', expression, 'skew', selection=selection, edges=edges)
+
+    sum_moment1 = _sum_moment(expression, 1, selection=selection, edges=edges)
+    sum_moment2 = _sum_moment(expression, 2, selection=selection, edges=edges)
+    sum_moment3 = _sum_moment(expression, 3, selection=selection, edges=edges)
+    count_ = count(expression, selection=selection, edges=edges)
+
+    m1 = sum_moment1 / count_
+    m2 = sum_moment2 / count_
+    m3 = sum_moment3 / count_
+    skew = (m3 - 3*m1*m2 + 2*m1**3) / (m2 - m1**2)**(3/2)
+    return skew