patch geeglm() error when starting values specified

[jeffeaton]

This PR patches issue #5 that geeglm() threw an error when argument start = was specified.

I also suggested one other small change. For the starting values of geese.fit(), I removed the condition:

  if (is.null(start)) 
    start <- glmFit$coef

So that it instead always uses the coefficients from glmFit as the starting values for geese.fit()

  start <- glmFit$coef

The specified starting values are still used as starting values for the initial glm() fit.

Maybe I am missing something though about why it reverts back to the initial starting values for initialising geese.fit().

Thanks,
Jeff