Bug fix release 2.3.5

What's new in 2.3.5

Version 2.3.5 is primarily a bug fix release. Here are the salient changes from 2.3.4:

• Fixed bug in invcdf unit test
• Generalized invcdf to accept n-dimensional argument
• Added bounds checking to Bernoulli
• Fixed failure in summary_plot with single chain
• Repaired tuning in GPStepMethod
• Fixed bug in HPD calculation
• Added optional chain argument to summary_plot
• Replaced rank with ndim to avoid numpy deprecation warning
• Fixed invalid variance in Geweke diagnostic
• Added statsmodels to pymc-requirements
• Fixed bug in Node.summary() for multidimensional nodes
• Fixed bad rounding in setting length of trace
• Changed constraint on rwishart and rwishart_cov to be n > p-1
• Added effective_n function for calculating effective sample size
• Tweaked BinaryMetropolis to assign it to Categorical variables where k=2
• Improved string formatting for Node.summary
• Changed tuning to prevent scale factor from shrinking to zero
• Changed default verbosity on gof_plot
• Improved Anaconda compatibility
• Fixed bug in batchsd for uneven batches