New stable build (0.22.0), with general linear/logistic regression + permutation

[Christopher Chang]

Features that have been added in the last four months include:
- General linear/logistic regression with permutation (--linear/--logistic).  (This is slow, but should not be unusably so.)
- Fast complete linkage clustering (--cluster).
- Classical multidimensional scaling (--mds-plot), with a list of eigenvalues if desired.
- Run-of-homozygosity analysis (--homozyg).  Pools of overlapping runs can also be analyzed (this is where the main speedups are).
- Support for arbitrary chromosome names (--allow-extra-chr).

As before, bugfixes will go into the stable 0.22.x builds, while new features will go into the 0.23+ development branch.

The next major stable build announcement, scheduled for early 2014, is planned to be a PLINK 1.50 release candidate.  It will include most remaining PLINK flags which have not been rendered obsolete by more accurate free software, including CNV analysis commands and multimarker tests.  (After discussing the issue with someone who actually spent some time trying to optimize MaCH, though, I have decided to just drop imputation for now.)  It will also add support for arbitrarily long allele names (which have actually been supported by PLINK this whole time).

The #1 priority after that point will be adding support for non-biallelic markers and phased data.  (This will be time-consuming.)

I will be in Boston for the ASHG 2013 meeting next week; you are welcome to talk to me there.