move models out of scipy.sandbox

In no particular order:

• Docstring listing functions and classes in info.py - see http://svn.scipy.org/svn/numpy/trunk/numpy/doc/DISTUTILS.txt
• tests for bssplines.py, bsplines_module.py
• Remove dependency on pylab for bsplines.py
• FIXMEs in cox.py, formula.py
• In bsplines.py: # note to self: check out eig_banded! in linalg.decomp?
• Docstrings! Note cox.py, at least

Requirements for stats routines inclusion in scipy that I guess we should follow: http://projects.scipy.org/scipy/scipy/wiki/StatisticsReview

The way the "namespace" parameter is meant to be used in various classes in formula.py is somewhat confusing. It seems that it sometimes takes on the value of formula.terms (a module level variable), sometime clobals() and other times has no default value.

What is the rationale behind this parameter and is there some way to refactor it to not rely on globals() or a module level variable?

DONE

• removed dependency on pylab in bspline.py 12/03/06
• FIXME in cox.py removed 12/17/06
• FIXME in formula.py removed 12/17/06
• removed "# note to self ..." bspline.py 12/03/06

IN PROGRESS

• Docstring listing functions and classes in info.py - see http://svn.scipy.org/svn/numpy/trunk/numpy/doc/DISTUTILS.txt
• tests for bssplines.py, bsplines_module.py
• Docstrings! Note cox.py, at least