add MonoidConcat class

[lexming]

This PR adds a variation of the Monoid class specifically optimized to concatenate lists.

Monoid is currently used by vsc-filesystems to concatenate lists of filesystem/quota attributes and it is painfully slow. To the point where some scripts such as dquota become unusable.

The issue is that the reduce method used in Monoid does not scale. The table below shows the time needed to inject into a MonoidDict an increasing amount of lists with the current mappend used in vsc-filesystems, some other alternative mappends and the new MonoidConcat class (without reduce):

1. Monoid with current mappend used by vsc-filesystems:

   l = lambda xs, ys: xs + ys

2. Monoid with alternative mappend with list.extend:

   def l(xs, ys):
       new_xs = list(xs)
       new_xs.extend(ys)
       return new_xs

3. Monoid with alternative mappend with list.append:

   def l(xs, ys):
       zs = []
       list(map(zs.append, xs))
       list(map(zs.append, ys))
       return zs

4. Monoid with alternative mappend with itertools.chain:

    l = lambda xs, ys: list(itertools.chain(xs, ys))

5. MonoidConcat without reduce

Nr of records	1: a + b (s)	2: extend (s)	3: append (s)	4: chain (s)	5: MonoidConcat (s)
100	0.0227	0.0469	0.0512	0.0502	0.0155
1000	0.3881	0.4753	1.9757	0.9320	0.1381
10000	30.4001	29.3562	171.8019	63.3293	1.3504
20000	122.6388	116.9318	--	--	2.3829

[lexming]

Not sure what happened with Jenkins, but running tox on my side works fine: https://gist.github.com/53be3bff4225c96598eb2ddcfe3889a7

[stdweird]

what is the full code for your timings? in particular why not use list(itertools.chain(*list_of_all_lists)) instead of using monid at all?

[lexming]

Superseded by https://github.com/hpcugent/vsc-filesystems/pull/84