ENH: support MultiIndex and tuple hashing

[jreback]

on top of #15216

closes #15227

[jreback]

cc @mikegraham

I separated this from the prior PR.

yes my goal here is to effectively emulate hash(tuple) but doing it efficiently (e.g. we already have the arrays in a packed format, turning into python objects, just to compute a tuple hash is very inefficient).

[jreback]

@mikegraham
ok I incorporated your commit (with a slight modification) and now passes tests!

[codecov-io]

Current coverage is 86.32% (diff: 100%)

Merging #15224 into master will increase coverage by 0.01%

@@             master     #15224   diff @@
==========================================
  Files           139        139          
  Lines         51096      51140    +44   
  Methods           0          0          
  Messages          0          0          
  Branches          0          0          
==========================================
+ Hits          44103      44148    +45   
+ Misses         6993       6992     -1   
  Partials          0          0          

Powered by Codecov. Last update be32852...8b1d3f9

[mikegraham]

There were a couple memory improvements in 187573b over the commit as you cherry-picked it, I think.

[jreback]

@mikegraham ahh ok, I will pick-again. thanks!

[jreback]

@mikegraham ok fixed.

Note that this obviously returns uint64, so it is not quite compat with the tuplehashing in python?

[mikegraham]

I stole the algorithm, but it's not compatible. I also iterate over the members in forward error rather than reverse...I don't think that is substantive.

[jreback]

cc @mikegraham any final comments?

@jorisvandenbossche

[mikegraham]

We might want to do something about this behavior.

>>> pandas.tools.hashing.hash_pandas_object(pandas.DataFrame(), index=False)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "pandas/tools/hashing.py", line 91, in hash_pandas_object
    h = _combine_hash_arrays(hashes, num_items)
  File "pandas/tools/hashing.py", line 19, in _combine_hash_arrays
    first = next(arrays)
StopIteration```

[jreback]

@mikegraham updated