Misc07 cleanup multipletesting and proportions

[josef-pkt]

first part: some cleanup of multipletesting,
see also http://jpktd.blogspot.ca/2013/04/multiple-testing-p-value-corrections-in.html
closes #763
closes #764
closes #766

Note: includes several backwards incompatible renames in weightstats

second part: statistical test for proportions and normal based tests

some functions still miss unit tests :

• (almost full coverage in weightstats (97%) and proportions (92%), one method and a few lines left untested)
• not all features tested, latest statistical tests don't have shape/2d tests

currently all written as individual functions :

• maybe add a class to collect methods, add plotting ?
• and more results classes
• the allpairs tests return a results instance, all others just numbers

some functions are still missing (ztest_prop, with more options than generic ztest), mostly done, I think

adding ztest and ztost to weightstats is incomplete (not attached as methods) DONE

ddof again:

• I need them for proportion and correlation for the z-based tests see b46ed52
• currently I added them to the standalone functions but not the methods

[josef-pkt]

this goes the wrong way, should be multitest_alias[a] = m[0]

[josef-pkt]

almost done here

proportion has essentially full coverage (but maybe not all options or use cases)

left to do:
integrating normal based tests, ztests, with weightstats classes is still missing,
and incomplete unittest coverage of ztest extra functions in weightstats.

[josef-pkt]

The last commit cd69578 added some new stuff that are not yet unit tested. Most of it is copy paste to finish up some loose ends and add generic versions, for example add the methods for ztest and ztost to weightstats.

I wrote those changes some time ago, and I'm not sure what the status is.
(I started tests for correlation, marginal homogeneity in contingency tables, and one sample outliers tests, but put yet in a new, not yet pushed, branch)

[josef-pkt]

Pretty much done here. The docstrings should get another review.

I did some interface consistency changes.

Some options still vary across functions and methods, but I leave that for the next round, which will also bring in tests for correlation, which are again based on normal distribution (ztest) after transforming the test statistic.
I opened issue #847 for that.

(Still bugging: it still uses confint_xxx instead of tconfint_xxx in parallel to zconfint_xxx )

[josef-pkt]

missing : add new functions and methods to api and sphinx docs

[josef-pkt]

I'm going to rebase, force push and then merge.

[josef-pkt]

travisCI python 2 fails again with python-dateutil, python 3.2 passes

merging